Fix/149 structural chunker drops documents with noheader

[flyingtony1424]

Issue

Closes #149

Changes

• ingestion/chunking/structural_chunker.py: _extract_sections() now appends content lines regardless of whether a heading has been seen yet, and always attempts to flush the trailing section (guarded by a non-empty-after-strip check so blank sections aren't emitted).
• tests/unit/test_issue_149_reproduction.py: reframed from "expected to fail" reproduction tests to permanent regression tests (docstrings/comments updated; no assertions changed) now that the underlying bug is fixed.
• JOURNAL.md: added Week 9 check-ins documenting progress.

Testing

• Unit tests pass (make test-unit)
• Integration tests pass (make test-integration) — not applicable, no integration surface touched
• Linter passes (make lint) — clean on the files changed in this PR; see note below on pre-existing repo-wide failures
• Type checker passes (make typecheck) — clean on the files changed in this PR; see note below
• New/updated tests cover the changes

How to manually verify:

1. Run the targeted regression tests: .venv/Scripts/python -m pytest tests/unit/test_issue_149_reproduction.py tests/unit/test_structural_chunker.py -v — all 18 should pass, including test_document_with_no_headings and the three issue-149 tests.
2. In a Python shell, confirm a heading-less document now produces a chunk instead of []:

   from ingestion.chunking.structural_chunker import StructuralChunker
   c = StructuralChunker()
   result = c.chunk("Plain paragraph text with no markdown headings at all.", {"source": "readme"})
   print(len(result), result[0].text)  # before the fix: 0, IndexError; after: 1, the full text

3. Confirm existing heading-based chunking is unchanged by running the full suite: .venv/Scripts/python -m pytest tests/unit -m unit -q and diffing against a run on main — the same 52 pre-existing failures should appear in both (see Notes below), with no new failures introduced by this branch.

Pre-existing failures observed (unrelated to this change):
Before making any changes, the local venv was missing several declared dev dependencies (redis, structlog, pypdf, python-jose, etc.); after running pip install -e ".[dev]", the full tests/unit suite has 52 pre-existing failures across unrelated modules (test_bias_detector.py, test_pii_scrubber.py, test_resume_parser.py, test_review_service.py, test_skill_extractor.py, test_tech_detector.py, etc.). I confirmed via git stash that these failures exist identically before and after this branch's changes — none touch structural_chunker or issue #149. Separately, ruff check has one pre-existing unused-variable warning in structural_chunker.py (current_level, present before this PR) that I left untouched as out of scope, and mypy fails repo-wide due to a numpy/Python 3.14 typestub incompatibility (Type statement is only supported in Python 3.12 and greater) unrelated to this change.

Screenshots / Demo

N/A — backend chunking logic change, no UI surface.

Notes for Reviewers

• The main design decision (flagged as a risk in PLAN.md): heading-less sections get heading_path="" (via " > ".join([])) and heading_level=0. I grepped rag/ and api/ for heading_path consumers and found none outside the chunking module itself, so an empty string should be safe — but flagging in case there's a downstream consumer I missed.
• Preserving preamble text means documents that previously had text before their first heading will now produce one additional chunk. This could shift chunk_index values for previously-ingested documents; may be worth a follow-up re-ingestion pass for existing indexed docs, but I didn't see an existing re-ingestion script wired to run automatically, so I left that out of scope for this fix.