Fix wily rank to reconstruct full project state at latest revision (#262)

[tonybaloney]

Fixes #262.

Problem

wily rank (without --revision) ranks against the most recent revision. But wily v2 stores metrics as deltas — a full "seed" revision plus only the files that changed in each subsequent revision. The previous implementation built its lookup from rows whose revision exactly matched the target revision:

for row in index:
    if row["revision"] == target_revision_key:
        revision_data[row["path"]] = row.get(resolved_metric.name)

So on any real repository, the latest revision contains only the file(s) changed in the last commit, and wily rank listed ~1 file by default. The no-path variant also surfaced directory/root aggregate rows that all shared an identical value.

Before (microsoft/graphrag, default latest revision)

$ wily rank packages/graphrag/graphrag raw.loc
┌────────────────────────────────────────────────────────────┬───────────────┐
│ File                                                       │ Lines of Code │
│ packages/graphrag/graphrag/index/utils/derive_from_rows.py │ 182           │
└────────────────────────────────────────────────────────────┴───────────────┘

After

$ wily rank packages/graphrag/graphrag raw.loc -l 10
┌─────────────────────────────────────────────────────────────┬───────────────┐
│ File                                                        │ Lines of Code │
│ packages/graphrag/graphrag/cache/__init__.py                │ 5             │
│ packages/graphrag/graphrag/callbacks/__init__.py            │ 5             │
│ ... (all files in the project)                              │ ...           │
└─────────────────────────────────────────────────────────────┴───────────────┘

Fix

In src/wily/commands/rank.py, reconstruct the project state at the target revision by taking, for each file path, the most recent row at or before the target revision's date — instead of exact-matching revision. Ranking is also restricted to path_type == "file" rows, so directory/root aggregates and function/class detail rows no longer pollute the table with duplicate values.

Tests

Added test_rank_latest_includes_unchanged_files, which builds a repo where src/stable.py exists only in the seed commit and src/churn.py is modified in a later commit, then asserts both appear when ranking the latest revision. (The existing gitdir fixture has a single file that changes in every commit, so it never exercised this bug.)

All test/integration/test_rank.py tests pass (16 passed); ruff check is clean.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 61.53846% with 5 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.00%. Comparing base (cd768bb) to head (f2f7307).
⚠️ Report is 1 commits behind head on master.

Files with missing lines	Patch %	Lines
src/wily/commands/rank.py	61.53%	2 Missing and 3 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #268      +/-   ##
==========================================
- Coverage   90.36%   90.00%   -0.36%     
==========================================
  Files          20       20              
  Lines        1193     1201       +8     
  Branches      161      163       +2     
==========================================
+ Hits         1078     1081       +3     
- Misses         87       89       +2     
- Partials       28       31       +3     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.