test: lift coverage on jvm-orphan-reaper (36->97%) and minimal-dap (85->92%)

[debugmcpdev]

Summary

• src/utils/jvm-orphan-reaper.ts: 36% → 97% line coverage. Expose parseArgs, listLinux, listDarwin, listWindows, defaultKill as @internal named exports so platform-specific helpers can be unit-tested on any host. Matches the file's existing test-seam pattern (lister/isAlive/killer are already injectable via ReapOptions). Add 31 tests covering pure parsing, process.kill paths, and all three platform listers via vi.mock of node:fs/promises + node:child_process.
• src/proxy/minimal-dap.ts: 85% → 92% line coverage. Add 10 tests for normalizeAdapterId branches, the non-stackTrace child-wait loop, child fallback (graceful completion, fall-through, rethrow), configurationDone deferral edge cases, and trace-file error swallowing.

No runtime behavior changes; only @internal export keywords added to previously-private helpers.

Coverage delta

File	Before	After	Lines missing
src/utils/jvm-orphan-reaper.ts	36%	97.32%	72 → 3
src/proxy/minimal-dap.ts	85%	92.20%	46 → 24

The reaper's remaining 3 missing lines are deep inside platform-specific error sub-branches that aren't worth contortion-testing. Minimal-dap's remaining 24 are inside deep child-session timing branches that would need integration-style fixtures rather than unit tests.

Test plan

• npm run lint — clean
• npx tsc --noEmit — clean
• npm run test:unit — 1758 / 1758 pass (+41 net new tests vs. baseline)
• npm run test:coverage:summary — overall coverage 87.72% statements / 88.2% lines
• npm run check:personal-paths — clean (pre-commit hook passed)
• Existing tests/unit/utils/jvm-orphan-reaper.test.ts (8 orchestration tests) still passes against the modified source

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!