fix(markdown): resolve optional token threshold defaults

[rudycelekli]

Problem

The public pageindex.md_to_tree() API defaults both optional token thresholds to None. Enabling summaries or thinning without explicitly supplying a threshold compares an integer token count against None and raises TypeError, before summaries can be generated.

Change

Resolve omitted or explicit None thresholds to the CLI's existing defaults: 200 summary tokens and 5,000 thinning tokens. Preserve explicit zero and positive thresholds.

Validation

• Reproduced both failures through the exported async API using real Markdown parsing and token counting before applying the fix (four failing subtests).
• Added coverage in the existing Markdown test file for omitted/None thresholds, preserved merged child text, and explicit zero/positive thresholds.
• Focused tests: 7 passed, 6 subtests passed. A falsy-default mutation fails both explicit-zero controls.
• Full local repository gate: python -m pytest -q — 589 passed, 218 skipped, 6 subtests passed on Python 3.12.14/PDFium 5.13.0, without optional agent frameworks.
• No live model/provider calls were made. Only the summary-generation boundary is stubbed where a model response is required.

The hosted Python 3.10/3.13 and optional-framework/PDFium matrix remains for upstream CI; the local result does not claim that matrix passed.

Prepared with AI assistance; source, reproductions and exact patch were independently reviewed.