fix(ingestion): allow leading whitespace before resume section headers

[pgazar]

Summary

Fixes resume section detection so it recognizes section headings (e.g. Education:, Skills:) even when they're preceded by leading whitespace (spaces or tabs). This is common in text extracted from PDFs and in indented resume text, and previously caused _detect_sections() to silently return an empty list.

Issue

Closes #147

Changes

• ingestion/parsers/resume_parser.py: _detect_sections() patterns now allow optional leading horizontal whitespace ([ \t]*, not \s*, so the match can't cross a blank line onto a header several lines below). Also removes the now-redundant \n-prefixed pattern variants, since re.MULTILINE already makes ^ match right after every newline.
• ingestion/parsers/resume_parser.py: fixed an unrelated pre-existing ruff B904 finding in the same function (_parse_pdf's except block) — needed to pass the project's pre-commit hook on this file.
• tests/unit/test_resume_parser.py: added 5 regression tests — tab-indented headers, mixed space/tab indentation, leading+trailing whitespace around a header, a guard test that indented body/bullet text merely mentioning a keyword (e.g. "skills") isn't misdetected as a header, and a check that headerless text still returns [].

Testing

Automated

• Unit tests pass (make test-unit / pytest tests/unit -m unit): 383 passed, 50 failed — the 50 are pre-existing failures unrelated to this change (see note below); before this change there were 53 (the 3 fixed here plus the same 50).
• Integration tests pass (make test-integration) — not run; this change doesn't touch integration surfaces
• Linter passes (make lint) on touched files (ingestion/parsers/resume_parser.py, tests/unit/test_resume_parser.py)
• Type checker passes (make typecheck) — ingestion/parsers/resume_parser.py passes cleanly; note below re: test file
• New/updated tests cover the changes

Manual verification

To see the bug and the fix directly, from the repo root run:

.venv/bin/python -c "
from ingestion.parsers.resume_parser import ResumeParser

# Note the leading spaces before 'Education:' and 'Skills:'
resume_text = '''
    John Smith
    john@example.com

    Education:
    - B.S. Computer Science

    Skills: Python
'''

parser = ResumeParser()
result = parser.parse(resume_text)
print(result.metadata['detected_sections'])
"

• Before this fix: prints [] (empty — both indented headings are missed)
• After this fix: prints ['Education', 'Skills']

You can also target just the previously-failing tests directly:

.venv/bin/pytest tests/unit/test_resume_parser.py -k "test_parse_single_column_resume_text or test_parse_resume_no_work_experience or test_detect_sections" -vv

All three should now pass (they failed on main before this PR).

Pre-existing failures (not introduced by this PR)

• make test-unit has 50 pre-existing unrelated failures on main (53 before this fix, since 3 of those were the resume-parser tests this PR fixes), spanning many unrelated modules (test_bias_detector.py, test_pii_scrubber.py, test_review_service.py, etc.). This PR does not add to that count.
• Two other test_resume_parser.py tests (test_parse_markdown_resume, test_strip_markdown_syntax) fail for an unrelated reason: _strip_markdown()'s markdown-header regex has the same "no leading whitespace" limitation this PR fixes for _detect_sections(), but for #-style markdown headers rather than resume section headers. Out of scope for Resume section detection fails on text with leading whitespace #147 (which is specifically about resume section detection); left untouched and unaffected by this PR either way.
• The repo's mypy pre-commit hook flags all 16 existing test functions in tests/unit/test_resume_parser.py (not just the 5 added here) for missing return-type annotations. This is a pre-existing gap across every file in tests/unit/ — make typecheck itself only checks api/ core/ ingestion/ rag/ agent/ safety/, not tests/, so this isn't part of the project's own declared "passes" bar. Left as-is to keep this PR scoped to Resume section detection fails on text with leading whitespace #147.
• make typecheck on main (baseline, unrelated to this PR) fails with 5 pre-existing errors: missing stubs for PyPDF2, jose, passlib.context, rank_bm25, and a numpy stub syntax error in the local venv. None are in ingestion/parsers/resume_parser.py.

Notes for Reviewers

Open to feedback on the whitespace-handling approach (e.g. [ \t]* vs. stripping each line before matching) if there's a preferred pattern elsewhere in the codebase. Also flagging the two pre-existing _strip_markdown() failures above in case a separate issue should be filed for those.