fix(ingest): normalize Markdown front-matter dates before storage

[Shraddhameduri]

Fixes #246

What

yaml.safe_load() converts unquoted YAML front-matter dates into datetime.date / datetime.datetime objects, which DynamoDB's TypeSerializer rejects at write time (TypeError: Unsupported type "datetime.date"). This PR normalizes parsed front-matter in the ingestion layer before it reaches storage.

Changes

• src/dynavec/ingest.py: new recursive _normalize_front_matter() helper (date/datetime → .isoformat(), recurses through dicts and lists, scalars untouched). Applied in MarkdownSource._front_matter(), which also covers the S3Source Markdown path since it reuses the same parser.
• tests/test_markdown_source.py: 5 new tests — unquoted dates → ISO strings, timestamps → ISO strings, nested dates in dicts/lists, quoted strings and scalars untouched, and an end-to-end check that ingested metadata passes boto3's TypeSerializer after _to_dynamo.

Test results

• pytest tests/test_markdown_source.py — 19 passed
• pytest tests/test_ingest.py tests/test_s3_source.py — 25 passed

[codeforstartups]

Correct bug fix for #246 — yaml.safe_load() turns unquoted front-matter dates into datetime.date/datetime objects that DynamoDB's TypeSerializer rejects. _normalize_front_matter recursively converts them to ISO-8601 strings (handling nested dicts/lists too). Verified locally: ruff clean, 19 markdown-source tests pass. Merging — thanks @Shraddhameduri! 🙌