Store commit-failure fakes raise inside the transaction body instead of at commit #1559

Problem

Two tests that claim to cover a failed store commit raise OSError("commit failed") inside the transaction body (dev a4357f9a):

• test_failed_atomic_control_commit_exposes_no_partial_transition (implementations/python/tests/test_run_310_supervisory_lifecycle.py:847-876) wraps _upsert_record and raises at :872. record_participant_control claims a running record (raes_runtime/participant_control_mediation.py:112) before it calls _commit_control_transition (:115), so the first _upsert_record call is the claim, and the test never reaches the atomic control commit it names.
• test_local_control_plane_store_rolls_back_snapshot_transaction_failure (implementations/python/tests/test_runtime_control_plane_api.py:1864-1887) wraps _upsert_snapshot and raises at :1878. It saves an empty RuntimeSnapshot(), which equals the default snapshot, so even a committed write would pass.

An error raised in the body takes the rollback branch of transaction() (raes_runtime/control_plane_store_local_codec.py:13-24). A real commit failure is a sqlite3 error raised by commit(): transaction() re-raises it without a rollback, and _connection() (raes_runtime/control_plane_store_local.py:357-362) closes the connection, which discards the open transaction. The audit observed this on SQLite 3.50.4 with a deferred foreign key: commit() raised sqlite3.IntegrityError, and nothing persisted after close().

The #1344 fixture-provenance audit lists this as finding ST-2, open, in the record #1461 proposes. The fix changes only tests, and #1472 proposes it.

Scope

• Make both fakes fail at SQLite's own COMMIT (for example through a deferred foreign-key violation in the open transaction) after the writes each test names, and assert on state that a committed write would change.

Done when

• Both tests fail if the store commits the transaction, and pass on dev.
• If docs(research): record the fixture provenance audit and the boundary fixture convention #1461 has merged, the ST-2 entry in docs/research/fixture-provenance/index.md and its row in inventory.md record the fix.

Related

• docs(research): record the fixture provenance audit and the boundary fixture convention #1461 (noted it: the fixture-provenance record it proposes lists finding ST-2, open) and Audit the test suite for fixtures that assert shapes the real system never produces #1344 (fixture-provenance audit).
• test(runtime): make the store atomicity fakes fail at sqlite's commit #1472 (proposes the fix, test only).