healthcheck: detect stale lab Julia env (installed Manifest vs pinned Manifest)

[Rchari1]

Problem

Found during live testing of the 1.1–1.3 stack (see the root-cause writeup on #72): the lab env at ~/.amico/julia had DirectTrajOpt 0.9.6 while the repo's pinned Manifest wants 0.9.7. The env predated the 0.9.7 bump and was never re-instantiated.

Result: every solve died at construction —

MethodError: no method matching IpoptOptions(; intermediate_callback::PulseEmitCallback)

(the solver-agnostic intermediate_callback on the Ipopt path is new in DTO 0.9.7) — so runs finished failed with no AMICODE_PULSE records and the inspector showed "no pulse data". healthcheck.mjs passed throughout: julia+project only checks that the project loads, not that it matches the pin. This is a silent demo-killer on any machine whose env predates a Manifest bump.

Proposal

Add a healthcheck item that compares the installed lab env against the repo/VSIX-pinned Manifest:

• Cheap textual pass: parse both Manifest.tomls and diff [[deps.*]] name→version (report the first few mismatches).
• Failure message should say the fix: re-run install.sh (which copies the pinned Manifest + Pkg.instantiate), or copy + instantiate manually.
• The pinned Manifest ships with the extension (packages/extension/julia/Manifest.toml), so the check needs no network.

Optionally install.sh could print a warning when overwriting an existing env Manifest that differs, but the healthcheck is the load-bearing guard.

Acceptance

• With the env deliberately rolled back to DTO 0.9.6, healthcheck.mjs fails with a message naming the stale package(s) and the fix.
• With the env in sync, the item passes.