Config: expose style traversal thread count

[HenriqueAnzoategui]

Summary

Adds an additive style_thread_count: Option<i32> field to Config so callers can disable stylo's parallel style traversal when invoking render concurrently from multiple OS threads.

Without this, two concurrent renders share stylo's process-global STYLE_THREAD_POOL (initialised by blitz). Rayon workers in that pool hold per-thread AtomicRefCell sharing caches borrowed mutably for each traversal's lifetime, and a second arrival panics with already mutably borrowed. This bites pretty much any embedder driving render() concurrently — servers rendering HTML per request, batch pre-renderers, even hyper-render's own integration test suite (see the --test-threads=1 workaround in PR 3's test summary).

None (default) preserves the historical auto-thread behaviour. Passing Some(1) keeps style work on the calling thread, where the thread-local cache is uniquely owned. Per-render parallelism within stylo is negligible at typical hyper-render document sizes; the real-world tradeoff is a small per-render latency bump in exchange for safe concurrency across renders.

Changes

• src/config.rs
  • New pub style_thread_count: Option<i32> field on Config with full rustdoc covering the rationale and the one-shot STYLE_THREAD_POOL initialisation gotcha.
  • New builder method pub fn style_thread_count(mut self, count: i32) -> Self with doctest.
• src/lib.rs — create_document plumbs config.style_thread_count into DocumentConfig.stylo_thread_count (blitz-side field name).

Compatibility

Fully additive: new field on a struct that already derives Default, new builder method that defaults to None. Every existing caller is unaffected.

Tests

• 1 doctest on the new builder method asserting the field round-trips.
• cargo test --all-features ✓ once the underlying blitz field is available.

Related work

• This PR is the second of a two-part series; the first half landed blitz's stylo_thread_count field (dioxuslabs/blitz#436). Together they let hyper-render callers configure stylo's parallel traversal behaviour without taking a direct dep on stylo internals.