test: restore basic_profile.json test fixture (#106)

[michellejtan]

Summary

Restores the missing tests/fixtures/sample_profiles/basic_profile.json test fixture referenced by scripts/run_evals.py and scripts/issues_manifest.json (#106). The fixture is a realistic fake portfolio (GitHub username, resume text, two repos) shaped to match what the ingestion parsers (ResumeParser, RepoAnalyzer, ReadmeParser) actually consume, so it's immediately useful if a future test wires it up rather than an arbitrary shape.

Issue

Closes #106

Changes

• Add tests/fixtures/sample_profiles/basic_profile.json — a fake profile with github_username, resume_filename, resume_text (using section headings like Summary/Technical Skills/Experience/Education so it exercises ResumeParser's section detection), portfolio_url, and exactly two repos entries shaped like GitHub API metadata (name, description, html_url, language, stargazers_count, forks_count, open_issues_count, pushed_at, readme_content with real markdown headings).
• Add a sample_profile_data fixture to tests/conftest.py that loads the JSON file, following the existing pattern of sample_resume_text/sample_readme_text.
• Add tests/unit/test_fixtures.py with a test proving the fixture loads correctly and has the expected shape (profile fields present, exactly two repos, each repo has name/html_url/readme_content).

Testing

• Unit tests pass (make test-unit)
• Integration tests pass (make test-integration) — not applicable, no integration tests reference this fixture yet
• Linter passes (make lint)
• Type checker passes (make typecheck)
• New/updated tests cover the changes

All checks passed locally:

• Unit tests: 53 pre-existing failures remain unchanged; new test in
  tests/unit/test_fixtures.py passes
• Integration tests: Not applicable
• Linting: No new errors introduced (182 pre-existing errors unchanged)
• Type checking: Passes

To manually verify:

1. Check out this branch and run pytest tests/unit/test_fixtures.py -v —
   confirm test_sample_profile_data_shape (or your actual test name) passes.
2. Open tests/fixtures/sample_profiles/basic_profile.json and confirm it
   contains a GitHub username, resume text with section headings (Summary,
   Technical Skills, Experience, Education), and exactly two repo entries
   with GitHub API-style fields (html_url, language, stargazers_count,
   etc.).
3. Run python -c "import json; d = json.load(open('tests/fixtures/sample_profiles/basic_profile.json')); print(d.keys())"
   to confirm the top-level keys match what sample_profile_data in
   tests/conftest.py expects.

Pre-existing failures (unrelated to this change)

Before starting, I ran make test-unit and make check on main to establish a baseline:

• make test-unit: 53 pre-existing failures (e.g. test_bias_detector.py, test_pii_scrubber.py, test_review_service.py, test_resume_parser.py, test_readme_parser.py, test_tech_detector.py, and others), 376 passing.
• make check: 182 pre-existing lint errors, none in files this PR touches.

I verified these are unrelated to this fix by stashing my changes and re-running both commands — the failure/error counts were identical with and without my changes applied.

After my changes:

• make test-unit: same 53 pre-existing failures, plus my new test passes.
• make check: still exactly 182 pre-existing lint errors — zero new errors introduced. tests/conftest.py and tests/unit/test_fixtures.py pass lint, format, and mypy individually.

This PR does not fix these pre-existing issues (out of scope for #106) but does not introduce any new failures either.

Screenshots / Demo

N/A — test fixture only, no UI change.

Notes for Reviewers

• No integration test currently consumes this fixture — scripts/run_evals.py has a TODO referencing tests/fixtures/sample_profiles/ for future benchmark portfolios, but wiring that up is out of scope here.
• I chose to shape the two repo entries like raw GitHub API metadata (html_url, stargazers_count, etc.) rather than matching the IngestedSource DB model, since that's the layer RepoAnalyzer.parse() and ingest_repo_metadata() actually consume before mapping into DB rows. Flagging in case a different shape is expected.