fix(runtime): preserve bootstrap failures when cleanup fails

[morluto]

Problem

If runtime bootstrap fails after opening its artifact store and ArtifactRepository.close() also raises, the cleanup exception replaces the original construction error. Users then see a secondary teardown fault instead of the startup root cause.

Fixes #676.

Solution

Keep cleanup best-effort inside the existing bootstrap exception path. When closing the store also fails, attach that secondary failure as an exception note and re-raise the original exception unchanged.

The regression closes a real store before injecting the cleanup error, so it proves both resource release and exception precedence rather than bypassing cleanup.

Testing

The regression failed on main by raising OSError: store close failure instead of the injected primary RuntimeError.

uv run --locked pytest -q tests/composition/runtime/test_runtime_lifecycle.py::test_bootstrap_cleanup_failure_preserves_primary_failure
uv run --locked ruff check src/jacobian/runtime/bootstrap.py tests/composition/runtime/test_runtime_lifecycle.py
uv run --locked ruff format --check src/jacobian/runtime/bootstrap.py tests/composition/runtime/test_runtime_lifecycle.py
git diff --check origin/main
make test-plan BASE=origin/main

Result: the regression passed and preserved the cleanup error in __notes__; Ruff, formatting, and diff checks passed. The planner conservatively selects broader product lanes for the transitively imported bootstrap path; those draft-PR checks are pending.

Trust & Compatibility Impact

No verification, checker authorization, artifact, protocol, or public API contract changes. Only dual-failure exception reporting changes; a cleanup-only failure is still raised normally.

Checklist

• Routine local validation passes (make check)
• Relevant focused tests are listed above

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.