fix(ingestion): capture structlog output in pytest caplog

[parker-cassar]

Summary

Wire structlog into stdlib logging for the test suite so pytest's caplog fixture can see application log events. This fixes suite-wide failures where structlog printed warnings to stdout but caplog.text stayed empty (notably test_empty_chunks_list_returns_empty).

Issue

Closes #159

Changes

• Add an autouse configure_structlog_for_caplog fixture in tests/conftest.py that configures structlog with LoggerFactory, stdlib processors, and ConsoleRenderer, then resets defaults after each test
• Add tests/unit/test_structlog_caplog.py covering warning and info capture via caplog

Testing

• Unit tests pass for changed behavior (test_empty_chunks_list_returns_empty and new structlog/caplog tests)
• Integration tests pass (make test-integration) — not required for this test-only change
• Linter passes on changed files (ruff check tests/conftest.py tests/unit/test_structlog_caplog.py)
• Type checker / pre-commit hooks passed on commit
• New/updated tests cover the changes

Pre-existing failures (unchanged by this PR)

make check and full make test-unit already fail on main for unrelated reasons. Before this change: 55 failed, 375 passed. After: 52 failed, 378 passed (the empty-chunks caplog failure plus the two new tests now pass; remaining failures are pre-existing in bias_detector, faithfulness_checker, pii_scrubber, review_service, etc.). Lint reports ~182 ruff errors and mypy reports ~103 errors across the codebase, none in the files this PR touches.

Screenshots / Demo

N/A (test infrastructure only)

Notes for Reviewers

• Production logging in core/logging.py is intentionally unchanged
• cache_logger_on_first_use=False so module-level structlog.get_logger() proxies pick up the test config
• Please focus review on whether the processor chain is the right test-only approach vs calling configure_logging() directly