Bump mojo-regex to 0.10.0

[msaelices]

Bumps mojo-regex from 0.9.0 to 0.10.0. Performance tuning release . Mojo vs Rust win rate improved from 57% to 64%.

Changes since 0.9.0 (all in msaelices/mojo-regex):

• #93: SIMD range comparison for contiguous byte ranges. range_lowercase 9.4x faster than Python.
• #94: DFA inner-loop optimization. 1.3-1.8x faster on DFA-heavy findall.
• #95: StringSlice text parameter through the matcher chain. ~11% average speedup.
• #98: NFAMatcher lazy DFA initialization fix (init_pointee_move).
• #99: StringSlice pattern parameter + hash-keyed regex cache.
• #102: Inline NFA per-character range checks (Dict lookup removal).
• #103: Implement re.sub() for regex pattern substitution.
• #104: Precompute range classification tag on ASTNode. Net -14% code in nfa.mojo.
• #105: Eliminate per-match String copies in DFA/NFA hot paths.
• #106: Add sub() comparison benchmarks. vs Python: 5/5 wins (1.2x-9.0x faster).
• #108: Capture group extraction and \1..`\9interpolation insub()`.
• #110: DFA fast path for sub() on fixed-width \d{N} capture groups. Skips NFA entirely, computes group boundaries via pointer arithmetic. Beats Python 2.8-3.2x and Rust 1.5-1.9x on phone number formatting.

Recipe changes:

• version: 0.9.0 -> 0.10.0
• rev: pinned to 127c69fe93981221d88a4ec8264265c901ad3221 (current main HEAD).

[msaelices]

Bumped the pinned rev to include msaelices/mojo-regex#98 (fix for a broken `ptr[] = value` init pattern in `NFAMatcher`'s lazy DFA allocation). The fix needs to be in 0.10.0 since it is a latent double-free introduced in 0.9.x's lazy DFA work and is the kind of thing that only bites at allocator-layout quirks.

Recipe now pins `rev: cbfdabc07804eccce24b19277b482072c6b969a1`.

[msaelices]

Bumped the pinned rev again to include msaelices/mojo-regex#99 (hash-keyed regex cache, pattern: StringSlice in the public API) and the updated CHANGELOG entry documenting both #98 and #99 under v0.10.0.

New pin: `rev: 6a8a0bd0e793527c5fc5acab881825b95a5da36b` (mojo-regex main HEAD).