feat(#59): harden integration snapshot test and add make docker-integration

[longieirl]

Closes #59 (local scope — CI changes deferred to a future issue).

Summary

• Hardens pytest -m integration with 5 new structural invariant assertions using data the pipeline already produces
• Adds make docker-integration for local Docker pipeline validation

Changes

test_output_snapshot.py — new _assert_output_invariants() function called after every snapshot comparison:

• 1a excluded_files.json: total_excluded == pdf_count - pdfs_extracted, each entry has reason + pages > 0
• 1b ibans.json: record count == pdfs_extracted, each masked IBAN matches [A-Z]{2}[0-9]{2}*...[0-9]{4}, IBAN suffix maps to output filenames
• 1c Transaction cross-check: sum of IBAN-specific CSV rows == transactions, CSV row count == JSON record count per IBAN pair, transactions > 0, pdfs_extracted > 0
• 1d CSV column integrity: no empty/duplicate columns, Date + Details + Debit/Credit present
• 1e Duplicate file structure: each duplicates*.json is a valid array, global duplicates.json length matches summary count

snapshots/output_snapshot.json — re-baselined

Makefile — docker-integration target: checks image exists, checks ./input has PDFs, runs container with read-only input mount, asserts non-zero CSV output

Type

• New feature
• Bug fix
• Breaking change
• Refactoring
• Documentation
• Performance
• Security

Testing

• Tests pass (coverage ≥ 91%)
• Manually tested

Checklist

• Code follows project style
• Self-reviewed
• Documentation updated (if needed)
• No new warnings