Chore/evaluator check cleanups

[lszhuhaichao]

Summary

This change continues the evaluator refactor by consolidating repeated path, baseline, substring, and regex-checking logic into shared helpers under evaluator/shared/check_utils.py, then applying those helpers across multiple evaluator checks.

What changed

• Extended shared evaluator utilities with reusable helpers for:

  • multi-root file scanning
  • regex pattern matching/counting
  • substring presence/counting
  • relative-path allowlist / top-level path validation
  • baseline path classification:
    • missing in submission
    • missing in baseline
    • created in submission
    • deleted from submission
    • modified relative to baseline
  • shared brace-matching extraction used by both class-body and function-body parsing

• Refactored multiple evaluator checks to use shared helpers instead of local ad hoc logic:

  • regex scanning loops
  • substring presence checks
  • file existence checks
  • path-boundary checks
  • baseline-protected file checks
  • Path.cwd() / CASE_REL style root resolution
  • local wrapper utilities like collect_code_files(...)

Key shared helpers introduced / expanded

• scan_files(..., suffixes=...)
• find_matching_patterns(...)
• find_missing_patterns(...)
• has_any_pattern(...)
• count_matching_patterns(...)
• has_any_substring(...)
• has_all_substrings(...)
• count_matching_substrings(...)