Extract test result CRUD out of the crud/__init__.py monolith

[akwasigroch]

Purpose

Next step in the incremental split of the crud monolith, continuing the series from #2410 through #2458. crud/__init__.py only shrinks from here, per the crud-layout rule in apps/backend/AGENTS.md.

Test Result was picked deliberately over the larger remaining blocks. Every branch currently in flight that touches crud/__init__.py — #2446 plus the local metric-tuning branches — modifies the same three spots: the import block, get_test_sets, and get_tests. Extracting TestSet or Test next would land right on top of them. Test Result sits below all three, needs no change to the import block, and has no cross-references from anywhere else in the module, so it lifts out cleanly. git merge-tree against #2446's branch reports no conflicts.

What Changed

• New app/crud/test_result.py holding get_test_result, get_test_results, create_test_result, update_test_result, delete_test_result, and the _TEST_RESULT_RELATED_FIELDS eager-load tuple. Bodies moved verbatim; no behaviour change.
• crud/__init__.py drops the 76-line Test Result section (1412 → 1336 lines). Its import block is untouched, so the line the in-flight branches add there stays clean.
• Callers switched to the house pattern used by the already-extracted modules, from rhesis.backend.app.crud import test_result as test_result_crud: routers/test_result.py (10 sites), tasks/execution/executors/results.py (2), tasks/execution/executors/output_providers.py (1), services/test_run.py (1).
• crud dropped from the top-level import in the three files where every use of it was a test-result call. services/test_run.py keeps it — it still reaches for test, prompt and test_configuration.
• Tests: 12 patch() targets retargeted from ...crud.get_test_result to ...test_result_crud.get_test_result, 4 direct calls updated, and 2 now-unused crud imports removed.

Additional Context

• Continues the split series: Extract trace CRUD out of the crud/__init__.py monolith #2410, Extract metric CRUD out of the crud/__init__.py monolith #2411, Extract project CRUD out of the crud/__init__.py monolith #2412, Extract comment CRUD out of the crud/__init__.py monolith #2438, Extract token CRUD out of the crud/__init__.py monolith #2439, Extract embedding CRUD out of the crud/__init__.py monolith #2441, Extract test run CRUD out of the crud/__init__.py monolith #2442, Extract tag CRUD functions into their own module #2449, Extract task CRUD functions into their own module #2450, Extract source CRUD functions into their own module #2451, Extract Behavior CRUD out of the crud/__init__.py monolith #2455, Extract File CRUD out of the crud/__init__.py monolith #2456, Extract Model CRUD out of the crud/__init__.py monolith #2457, Extract User CRUD out of the crud/__init__.py monolith #2458.
• Pure refactor — no schema, endpoint, or query changes. The generated SQL is identical.
• Remaining candidates after this one, in rough order of cleanliness: Organization (drags get_session_variables along), Tool, Architect session/message. TestSet and Test should wait until the metric-tuning work has landed.

Testing

cd apps/backend
uv sync --extra all --extra ee
uv run pytest ../../tests/backend/tasks/ ../../tests/backend/crud/ ../../tests/backend/routes/ ../../tests/backend/auth/ ../../tests/backend/services/ ../../tests/backend/security/ ../../tests/backend/metrics/

4520 passed, 42 skipped, 1 xfailed locally — that covers every test that touches test results, including the retargeted mock patches, which would fail loudly with AttributeError if a patch target had gone stale. Ruff check and format are clean on all touched files.