feat(test-coverage-reporter): token optimization — ~57% reduction per run (#4945)

* Initial plan

* feat: optimize test-coverage-reporter workflow for token efficiency

- Remove coverage-json step (duplicate of coverage-table, saves ~150K-300K tokens/run)
- Add discussion-template pre-build step (saves ~400K-600K tokens/run)
- Simplify agent prompt to only write Notable Findings + Recommendations
- Remove duplicate CRITICAL_GAPS and COVERAGE_TABLE injections from agent prompt
- Fix FUNC_AUDIT ternary regex to use branch counts instead of misleading line listing
- Add paths filter to push trigger to reduce run frequency on non-source changes
- Update workflow tests to cover all new token optimization assertions

* fix: double-quote expression vars, update description, add expansion test

- Switch single-quoted ${{ }} expressions to double quotes so
  GH_AW_EXPR_* variables expand at runtime in the compiled lock file
- Update description to reflect push.paths filter (not 'every push')
- Add test assertion verifying echo lines use double-quoted expansion
- Recompile lock file (now has --build-local and --session-state-dir)

Addresses all 8 Copilot review threads on PR #4945.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: Landon Cox <landon.cox@microsoft.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Author: 3 people