rust(feat): bulk annotation updates

[evan-sift]

Description

update_annotation now updates one or more annotations in a single call. It takes a required annotation_ids list (1 to 1000 ids) and applies the same change set to every id.

Archiving goes through one BatchArchiveAnnotations request. Unarchiving clears each annotation's delete date through per-annotation UpdateAnnotation requests with a bounded fan-out, the mechanism the API documents for restoring an annotation (the dedicated unarchive RPCs are not implemented by the service). Field changes are likewise one UpdateAnnotation per annotation with the same fan-out. When is_archived is combined with field updates, the updates run first, then the archive or unarchive.

Failure handling is explicit rather than all-or-nothing: per-annotation failures come back in failures with the id, code, and message next to the successful results, a backend-wide failure stops later batches and reports the untouched ids in not_attempted, and any partial failure sets the tool result's isError flag. next_step tells the agent which ids are safe to retry.

Stack note: the bundled agent skill and the 0.5.0 changelog for this whole stack land at the tip (#763).

Verification

• cargo test -p sift_mcp on this branch: 430 passed, 0 failed. New tests cover batch construction, per-id failure mapping, not_attempted reporting on backend-wide failure, archive ordering with combined field updates, bounded fan-out concurrency with order preservation, and parameter validation (empty list, empty ids, over-limit list, no updatable field).
• cargo fmt --all -- --check: clean.
• Live end-to-end against a development org (2026-08-26): bulk update, batch archive, per-annotation unarchive (verified both directions on two annotations, zero failures), and the create gate all exercised through an MCP client driving the built server.
• Rebased into the MCP tool stack (base rust/mcp-user-defined-functions).