fix(context): report unready graphs

[HsiangNianian]

Linked issue

Closes #734

What & why

get_minimal_context_tool opened GraphStore before checking graph readiness. In a
cold worktree, that created an empty database and returned status: ok, which could
make an agent trust analysis that contained no graph data.

This change resolves graph paths without filesystem side effects before opening the
store and returns an actionable not_ready response for missing, empty, or
commit-stale graphs. The normal build path keeps its existing directory creation and
legacy migration behavior.

Regression coverage includes linked worktrees, default and external data directories,
registry overrides, legacy databases and sidecars, empty graphs, and stale build SHAs.

How it was tested

uv run pytest tests/ --tb=short -q --cov=code_review_graph --cov-report=term-missing --cov-fail-under=65
uv run ruff check code_review_graph/
uv run --with mypy --with types-networkx mypy code_review_graph/ --ignore-missing-imports --no-strict-optional
uv run --with bandit bandit -r code_review_graph/ -c pyproject.toml -q

• 2090 passed, 5 skipped, 2 xpassed
• 80.67% coverage

Checklist

• Tests added for new functionality
• All tests pass: uv run pytest tests/ --tb=short -q
• Linting passes: uv run ruff check code_review_graph/
• Type checking passes: uv run mypy code_review_graph/ --ignore-missing-imports --no-strict-optional
• Lines are at most 100 characters
• Docs updated where behavior changed (README, docs/, docstrings)

[github-actions]

code-review-graph review

Overall risk: 0.60 (MEDIUM) — 17 changed function(s)/class(es), 0 affected flow(s), 5 test gap(s)

Risk-scored changes

Risk	Level	Symbol	Location	Tested
0.60	medium	code_review_graph/tools/_common.py::graph_provenance	code_review_graph/tools/_common.py:59	no
0.47	medium	code_review_graph/main.py::get_minimal_context_tool	code_review_graph/main.py:187	yes
0.40	medium	code_review_graph/tools/context.py::_not_ready	code_review_graph/tools/context.py:17	no
0.30	low	tests/test_tools.py::TestGraphProvenance.test_missing_graph_database_has_no_envelope	tests/test_tools.py:2091	(test)
0.25	low	tests/test_tools.py::TestGetMinimalContext.test_mcp_wrapper_reports_missing_graph_without_creating_state	tests/test_tools.py:1824	(test)
0.15	low	code_review_graph/incremental.py::get_data_dir	code_review_graph/incremental.py:305	yes
0.15	low	code_review_graph/incremental.py::get_db_path	code_review_graph/incremental.py:355	yes
0.15	low	code_review_graph/tools/context.py::get_minimal_context	code_review_graph/tools/context.py:48	yes
0.05	low	tests/test_incremental.py::TestGetDbPath	tests/test_incremental.py:150	no
0.05	low	tests/test_incremental.py::TestGetDbPath.test_read_only_resolution_does_not_create_migrate_or_clean	tests/test_incremental.py:179	(test)

Test gaps

• code_review_graph/tools/_common.py::graph_provenance (code_review_graph/tools/_common.py:59)
• code_review_graph/tools/context.py::_not_ready (code_review_graph/tools/context.py:17)
• tests/test_incremental.py::TestGetDbPath (tests/test_incremental.py:150)
• tests/test_tools.py::TestGetMinimalContext (tests/test_tools.py:1771)
• tests/test_tools.py::TestGraphProvenance (tests/test_tools.py:1945)

Token savings: this graph-backed report used ~57,144 fewer tokens (~95%) than reading every changed file in full (estimated, chars/4 approximation).

---

Powered by code-review-graph — local-first analysis; no code leaves the CI runner.

[tirth8205]

After refreshing onto current main, the exact head exposed a real semantic conflict with merged #710: get_data_dir() still imported the removed _REGISTRY_PATH, so registry overrides were silently ignored, and the new registered-data test patched a symbol that no longer exists. I pushed cbe61f9 to use the dynamic default_registry_path() and isolate the test through CRG_HOME. Validation on the corrected head: 201 tools/incremental tests passed (1 skipped), 62 registry/daemon-path tests passed, real multiprocessing parity passed, Ruff/diff passed, and a 243-file graph build completed with zero errors. Waiting for the refreshed hosted matrix before merge.