fix(core): use Observation.permalink in build_context to match the search index

[groksrc]

Fixes #929

Summary

build_context rebuilt the synthetic observation permalink inline:

generate_permalink(f"{primary_item.permalink}/observations/{obs.category}/{obs.content}")

with no 200-char truncation (#446) and no content digest (#931), so for observations longer than 200 chars it returned permalinks that don't exist in the search index.

As suggested in the issue: use the Observation.permalink model property instead, so there is exactly one place that defines the synthetic observation permalink format. No lazy-load hazard — ObservationRepository.get_load_options() already eager-loads Observation.entity, and execute_query applies those options to the find_by_entities query that feeds build_context.

Test plan

• New regression test test_build_context_observation_permalinks_match_search_index: a >200-char observation's permalink from build_context must equal both the model property and the indexed row's permalink. Verified it fails against the previous inline construction.
• tests/services/test_context_service.py — 13 passed (SQLite), 12 passed / 1 pre-existing skip (Postgres via testcontainers)
• test-int/mcp/test_build_context_validation.py + test_build_context_underscore.py — 8 passed
• ruff check, ruff format --check, ty check src tests test-int — clean

Not included

The issue's second item — release notes for the #931-shipping version mentioning that pre-existing search_index rows keep digest-less permalinks until re-sync or bm reindex --search — belongs to the release runbook, not this change.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits.
Repo admins can enable using credits for code reviews in their settings.