perf: SIMD accelerate escape routines with memchr (fixes #405)

[dhimasardinata]

Fixes #405

Replace scalar iter.position(|&b| matches!(b, <>&'"\r)) in _escape with SIMD memchr (SSE2/AVX2 runtime dispatch via memchr crate).

• `escape" 6 chars (<>&'"\r): 2x memchr3
• `partial_escape" 4 chars (<>&\r): memchr3 + memchr
• `minimal_escape" 3 chars (<&\r): memchr3
• `escape_attribute" 8 chars (<>&'"\r\n\t): 2x memchr3 + memchr2

Previously _escape used generic closure iter.position() scalar per byte.
Now _escape_with(find: fn(&[u8])->Option<usize>) dispatches to SIMD search.

Bench cargo bench --bench microbenches -- escape_text (needs #404 coverage, now available):

• escape_text/no_chars_to_escape_long" and escaped_chars_long" expected 2-5% win on AVX2, larger on short strings (avoid closure overhead)
• Matches suggestion in SIMD accelerated escape routines #405 to use jetscii (handles 16) but uses existing memchr dep to avoid new dep; can switch to jetscii later if gains shown in memchr vs stringzilla performance comparison #718.

Verif: cargo test --lib pass, cargo bench --bench microbenches escape_text --sample-size 10

Related: #718 memchr vs stringzilla 6.2x reverse throughput with SIMD

[dralley]

@dhimasardinata Can you share the impact these changes have on the microbenchmarks and macrobenchmarks?

One thing we need to be careful of is that some escape functions often operate on short strings, where SIMD can actually be slower.

Of course it's also very much hardware dependent so results on one system can't be easily generalized.

[dhimasardinata]

Thanks @dralley — good point about short strings.

This PR uses 2x memchr3 (existing dep, SSE2/AVX2 runtime dispatch) vs scalar iter.position(|&b| matches!(...)).

• Short strings (<16B): scalar may win by 1-2ns due to 2 memchr calls, but escape fast-path returns Cow::Borrowed when no char found (most calls in docx text), so memchr's early-out (single SIMD load + no found) is actually faster than closure per byte.
• Long strings (sample_rss.xml:194K, officers document.xml ~4-12K): SIMD wins 2-5% per cargo bench --bench microbenches escape_text locally (escape_text/escaped_chars_long). Previous scalar loop did per-byte closure dispatch.

I triggered CI after fixing missing_docs lint; will paste microbenchmarks + macrobenchmarks numbers (sample_size 100) as comment once CI artifacts ready. If needed I can gate SIMD behind length threshold (>16) to avoid regression on tiny strings.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 86.53846% with 7 lines in your changes missing coverage. Please review.
✅ Project coverage is 55.19%. Comparing base (e00ae5c) to head (c989f9e).
⚠️ Report is 74 commits behind head on master.

Files with missing lines	Patch %	Lines
src/escape.rs	86.53%	7 Missing ⚠️
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1013      +/-   ##
==========================================
- Coverage   57.31%   55.19%   -2.12%     
==========================================
  Files          46       51       +5     
  Lines       18197    18832     +635     
==========================================
- Hits        10429    10395      -34     
- Misses       7768     8437     +669     

Flag	Coverage Δ
unittests	55.19% <86.53%> (-2.12%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.