feat(checks): let OrphanExport report exports imported only by tests (CD-320)

[TAJD]

Closes CD-320. Read the correction below before the diff — the ticket's premise was wrong, and the finding that replaces it is filed as CD-325.

The ticket said

Design.OrphanExport drops test-file import edges, so an export a test imports is reported as "never imported in the project". Evidence: all 33 findings on poker-puzzle at 0.4.2 are symbols that tests import.

What is actually true

cofferdam counts a test-file import like any other. Minimal reproduction:

apps/worker/src/lib/account-capabilities.ts   export function entitlementFor(x: number): number { … }
apps/worker/test/account-capabilities.test.ts import { entitlementFor } from '../src/lib/account-capabilities'

cofferdam check . --no-cache at 0.4.2 → clean. Multi-line and type-mixed specifier lists behave the same. Add a one-line .cofferdamignore containing **/*.test.ts and the orphan finding appears.

That is what happened on poker-puzzle: its .cofferdamignore excludes **/*.test.ts and **/*.test.tsx, so the test files contributed no edges. An ignore rule silently changes the answers other files get, and Design.OrphanExport states that answer as a fact about the whole project. Filed as CD-325; not fixed here.

What this PR does ship

The capability the ticket asked for was genuinely missing, so it is kept:

• test_imports_count, a bool defaulting to true — the pre-existing behaviour, now named and switchable.
• FileConsumption tallies consumption twice per file: any counts every importer, non_test counts only importers that miss test_file_patterns. Classification reuses the existing pattern list rather than inventing a second definition of "test file".
• Under test_imports_count = false, an export consumed solely by tests reports "imported only by test files" — distinct from the genuine "never imported" case, which still reports in both modes.
• The namespace-import shortcut (a namespace import claims every named export) moves into is_consumed so both tallies get it identically, replacing the old continue special case.

Before this, "exported solely for tests" — a real smell — could not be seen at all.

Tests

Six cases: same-directory .test.ts and __tests__/ conventions, each in both modes; a normal-file importer (never flagged either way); no importer at all (flagged "never imported" both ways). Also fixes the pre-existing opts_default() helper in design/mod.rs for the new field.

Verification

cargo build --workspace, cargo test --workspace (1382 passed, 8 ignored), cargo clippy --workspace --all-targets -- -D warnings, cargo fmt --all -- --check — all green. cofferdam gen-docs re-run and committed.

🤖 Generated with Claude Code

https://claude.ai/code/session_019afugU2bTYCtXn6Vy9dn1v

[github-actions]

Criterion benchmark comparison (PR head vs base branch)

�[1m�[92m   Compiling�[0m cofferdam-checks v0.4.2 (/home/runner/work/cofferdam/cofferdam/crates/cofferdam-checks)
�[1m�[92m   Compiling�[0m cofferdam-engine v0.4.2 (/home/runner/work/cofferdam/cofferdam/crates/cofferdam-engine)
�[1m�[92m    Finished�[0m `bench` profile [optimized] target(s) in 49.34s
�[1m�[92m     Running�[0m benches/engine_bench.rs (target/release/deps/engine_bench-e0ce420cbfc0fb1d)
Gnuplot not found, using plotters backend
Benchmarking full_run_no_cache
Benchmarking full_run_no_cache: Warming up for 3.0000 s
Benchmarking full_run_no_cache: Collecting 20 samples in estimated 7.1123 s (420 iterations)
Benchmarking full_run_no_cache: Analyzing
full_run_no_cache       time:   [16.532 ms 16.679 ms 16.909 ms]
                        change: [+0.6774% +1.9444% +3.5728%] (p = 0.01 < 0.05)
                        Change within noise threshold.
Found 2 outliers among 20 measurements (10.00%)
  2 (10.00%) high severe

Benchmarking single_file_edit_incremental
Benchmarking single_file_edit_incremental: Warming up for 3.0000 s

Warning: Unable to complete 20 samples in 5.0s. You may wish to increase target time to 9.6s, enable flat sampling, or reduce sample count to 10.
Benchmarking single_file_edit_incremental: Collecting 20 samples in estimated 9.5992 s (210 iterations)
Benchmarking single_file_edit_incremental: Analyzing
single_file_edit_incremental
                        time:   [7.3557 ms 7.4005 ms 7.4598 ms]
                        change: [-3.4150% -2.4237% -1.4780%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 5 outliers among 20 measurements (25.00%)
  2 (10.00%) low mild
  2 (10.00%) high mild
  1 (5.00%) high severe