Merge TASK-053: cut dispatch over to lookup_v2 and retire v1 resolver

Retires the v1 registered_resources* HTTP dispatch path. After this
merge resolve_resource_for_request calls lookup_v2() exclusively;
the v1 registered_resources* maps survive only as registration-side
bookkeeping (lambda/class conflict detection, WebSocket dispatch).

Includes the validation iter1 cleanup: cache_key move semantics,
find_by_view bucket-iterator promotion, std::regex::nosubs, bench
path hoisting, wait_for_server_ready test helper, is_prefix
assertion on the 405 path, and the cache_find_by_view_promotes_to_front
LRU regression test.

Conflicts (all resolved by keeping the post-cutover semantics):
- src/httpserver/detail/route_cache.hpp: TASK-056 had added a
  drive-by empty-cache early-out anticipating this merge; kept
  TASK-053's wording + mutable bucket iterator required by the
  iter1 promotion fix.
- src/detail/webserver_dispatch.cpp: same defensive-copy hunk;
  kept TASK-053's wording and the route_cache_v2 -> route_lru_cache
  rename.
- specs/architecture/04-components/route-table.md: combined
  TASK-053 implementation status with TASK-056's CWE-407 note.
- specs/tasks/v2-deferred-backlog-plan.md: kept the checked-off
  TASK-053 action items; merged the TASK-056 2x radix-regression
  budget verification note into the bench_route_lookup entry.
- test/Makefile.am: kept TASK-053's v2_dispatch_contract entry
  plus the auth_handler_* entries that landed after the branch
  was cut.

Author: etr

specs/architecture/04-components/route-table.md

@@ -27,6 +27,6 @@ A `route_entry` carries:
 
 **Related requirements:** PRD-HDL-REQ-002, PRD-HDL-REQ-004, PRD-HDL-REQ-006.
 
-**Implementation status:** TASK-025 introduced `detail::route_entry` and the `lambda_resource` shim into the existing v1 three-map storage shape. TASK-027 wired `route_entry` into the full 3-tier table described above (hash map for exact paths, radix tree for parameterized/prefix paths, regex chain for regex routes). As of TASK-027 all three tiers are operational and the v1 three-map shape is maintained in parallel for backward-compatible dispatch until the v1 path is fully retired. TASK-056 swapped the radix-node child container from `std::unordered_map` to `std::map<…, std::less<>>` for CWE-407 immunity and added registration-time detection of prefix-vs-exact terminus collisions (`reject_terminus_collision`).
+**Implementation status:** TASK-025 introduced `detail::route_entry` and the `lambda_resource` shim into the existing v1 three-map storage shape. TASK-027 wired `route_entry` into the full 3-tier table described above (hash map for exact paths, radix tree for parameterized/prefix paths, regex chain for regex routes). TASK-053 retired the v1 dispatch path: `webserver_impl::resolve_resource_for_request` now consults `lookup_v2()` (cache → exact → radix → regex) directly, the v1 LRU cache and the four v1 lookup helpers (`lookup_route_cache`, `scan_regex_routes`, `store_route_cache`, `apply_extracted_params`) are deleted, and the LRU cache field is renamed `route_lru_cache`. The v1 registration-side maps (`registered_resources`, `registered_resources_str`, `registered_resources_regex`) survive *only* as registration-time bookkeeping for `prepare_or_create_lambda_shim` (lambda/class conflict detection) and the WebSocket dispatch path, both of which are non-HTTP-dispatch concerns; their removal is its own follow-up task. TASK-056 swapped the radix-node child container from `std::unordered_map` to `std::map<…, std::less<>>` for CWE-407 immunity and added registration-time detection of prefix-vs-exact terminus collisions (`reject_terminus_collision`).
 
 ---

specs/tasks/v2-deferred-backlog-plan.md

@@ -42,6 +42,7 @@ on `canonicalize_lookup_path`) can ride along with TASK-053.
 **Milestone:** M5 - Routing, Lifecycle, Builder & Features
 **Component:** `webserver_impl` dispatch
 **Estimate:** L
+**Status:** Done
 
 **Goal:**
 Make TASK-027's 3-tier route table (`exact_routes_`, radix tree, regex
@@ -52,30 +53,48 @@ the v1 vector. The architectural goal of TASK-027 (O(1) exact lookup +
 LRU cache + radix scan) is not realised end-to-end.
 
 **Action Items:**
-- [ ] Locate the v1 lookup site (`webserver_impl::find_route_for_request`
-  or equivalent in `src/detail/webserver_dispatch.cpp`) and add a feature
-  flag (`use_lookup_v2_`, default `true`) that selects `lookup_v2` over
-  the legacy walk.
-- [ ] Wire the `lookup_result` shape (tier_hit, captured_params) into the
-  existing dispatch contract so the call site doesn't need a parallel
-  `route_entry*` path.
-- [ ] Remove the v1 fallback walk once `make check` and
-  `routing_regression_test` pass under the v2 path.
-- [ ] Delete the legacy linear `route_table_v1_` field when no caller
-  remains. (May require a follow-up grep sweep of `webserver_register.cpp`
-  / `webserver_routes.cpp`.)
-- [ ] Run `test/bench_hook_overhead.cpp` and a new
-  `test/bench_route_lookup.cpp` to confirm the cache-hit path is in the
-  expected 100ns ballpark and the radix tier is in the µs range. When
-  creating `bench_route_lookup.cpp` also verify the TASK-056 acceptance
-  criterion: the `std::map`-based radix node container introduced by
-  TASK-056 must show no regression worse than 2× on the cache-miss radix
-  path compared to the former `std::unordered_map` baseline (the 2× budget
-  was formally deferred from TASK-056 to this task because `lookup_v2` was
-  not yet wired into dispatch when TASK-056 landed).
-- [ ] Drop the "TODO(Cycle K): rename route_cache_v2 → route_lru_cache"
-  comment in `webserver_impl.hpp:202` and do the rename now that v1 is
-  gone.
+- [x] Locate the v1 lookup site (`webserver_impl::resolve_resource_for_request`)
+  and add the safety net first: a unit-level contract test
+  (`test/unit/v2_dispatch_contract_test.cpp`) pinning the
+  end-to-end (method, path) → entry shape against `lookup_v2`
+  parity so step-2 cutover cannot regress silently. Committed
+  as TASK-053 step 1.
+- [x] Wire the `lookup_result` shape (tier_hit, captured_params) into
+  the dispatch contract: `resolve_resource_for_request` now
+  consults `lookup_v2()` directly behind the same `bool found,
+  route_entry entry` shape the v1 path expected, gated by the
+  `use_lookup_v2_` flag (default `true`). Committed as TASK-053
+  step 2.
+- [x] Remove the v1 fallback walk and rename `route_cache_v2` →
+  `route_lru_cache` in one commit so the dispatch path has a
+  single, named cache field. The four v1 lookup helpers
+  (`lookup_route_cache`, `scan_regex_routes`, `store_route_cache`,
+  `apply_extracted_params`) are deleted; the v1 registration
+  maps (`registered_resources*`) survive only as registration-
+  time bookkeeping for `prepare_or_create_lambda_shim` and the
+  WebSocket path, flagged in comments and the architecture doc
+  as non-dispatch state with their removal as its own follow-
+  up. Committed as TASK-053 step 3.
+- [x] Flip basic-suite integration tests whose v1 expectations no
+  longer match v2 semantics (`overlapping_endpoints`,
+  `regex_matching_arg_custom`, `regex_url_exact_match`) and
+  point them at `test/REGRESSION.md` §3/§4 plus the pinned
+  unit-level routing_regression tests. Committed as TASK-053
+  step 4.
+- [x] Add `test/bench_route_lookup.cpp`: pins cache-hit median ≤ 200
+  ns and 8-segment radix tier median ≤ 5 µs via
+  `webserver_test_access::impl(...)->lookup_v2(...)` (no MHD
+  daemon). Wired into `make bench` via `bench_targets`, NOT
+  `make check`. Sanitizer builds skip with exit 0. Committed
+  as TASK-053 step 5. The TASK-056 2× radix-regression budget
+  was also verified here per the acceptance-criterion deferral
+  from TASK-056.
+- [x] The "TODO(Cycle K): rename route_cache_v2 → route_lru_cache"
+  comment was removed when the rename landed in step 3; the
+  field is now `route_lru_cache` and the only surviving
+  `route_cache_v2` token in the headers is the historical
+  pointer comment ("Renamed from route_cache_v2 in TASK-053
+  step 3...") on the field declaration.
 
 **Dependencies:**
 - Blocked by: TASK-027 (Done), TASK-028 (Done)