fix(models): don't crash to_markdown() on a malformed heading level

[afonsojanu]

Fixes #176

Summary

ScrapeResult._item_to_markdown() sliced heading["level"][1:] and int()'d it unconditionally (int(heading.get("level", "h2")[1:])), so any level not shaped like h1-h6 raised ValueError ("title", "") or TypeError (None) instead of degrading gracefully. TextExtractor only ever emits h1-h6, but ScrapeResult also accepts user-supplied data (a documented, supported use), where one malformed heading crashed the entire to_markdown() call rather than just that heading.

Fix

Added _heading_level(), which parses the level defensively — falls back to h2 for anything not shaped h<digits>, and clamps in-range-shaped-but-out-of-bounds values (h9, h0) to 1-6 — and routed the heading loop through it, per the fix sketched in the issue.

Testing

• test_to_markdown_handles_a_malformed_heading_level — parametrized over "title", "", None, "h9", "h0": each renders without raising, with a heading line carrying 1-6 #s.
• test_to_markdown_clamps_out_of_range_heading_levels — pins the exact clamped output for h1, h6, h9, h0.
• Confirmed both fail against the old code (git stash on models.py, re-ran): 7 of 9 new cases failed with the exact ValueError/TypeError/assertion mismatches the issue describes.

uv run pytest tests/ -q
# 535 passed, 9 skipped, 19 deselected
uv run ruff check src/pyscrappy/core/models.py tests/test_core/test_models.py
# All checks passed!

mypy src/pyscrappy/core/models.py reports 4 pre-existing errors (missing pandas/yaml stubs, two no-any-return findings on unrelated lines) — none touch _heading_level or the lines I changed; unrelated to this fix.

[vedaant00]

Thanks for this @afonsojanu , and it's a clean fix, the _heading_level helper is nice. Two PRs landed for #176 within a minute of each other though, and I went with #177 since it was green and conflict-free. No reflection on this one. Really appreciate you jumping on it, especially right after #172, if you want another, #166 (bound the on-disk cache) is open and meatier.