Fix sticky+global [Symbol.match] .NET fast path returning wrong results

[lahma]

What

Fixes the sticky+global [Symbol.match] .NET fast path returning wrong results:

"aaa".match(/a/gy)     // was ["a","a"]      → now ["a","a","a"]
"ababab".match(/ab/gy) // was ["ab"]         → now ["ab","ab","ab"]
"aab".match(/a*/gy)    // was ["aa"]         → now ["aa","",""]

Why

The sticky branch in RegExpPrototype.Match compared match.Index (a string position) against
++li (a match counter), and called SetLength(li) after having written array index li:

• any match wider than one character made the position ≠ counter comparison break the loop after the
  first match, and
• even for single-character matches the final length truncated the last written element.

How

The loop now tracks the expected continuation position — previous match end, or +1 for empty matches
(AdvanceStringIndex; this fast path is guarded !fullUnicode, so the advance is always one UTF-16
unit) — and sets the array length to the number of matches written. lastIndex semantics are already
covered by the rx.Set(lastIndex, +0) at the top of Match (final observable value matches the spec's
failing-exec reset).

Tests

Test262 doesn't cover sticky+global [Symbol.match] through the .NET engine path, so this adds
regression tests (each shape runs against both the .NET fast path and the generic exec loop via the
u flag):

• adjacent single- and multi-char matches collected fully
• termination at the first non-adjacent position (later matches must not be included)
• empty-match advancement (/a*/gy)
• lastIndex reset to 0 after the call

Gating

• dotnet build Jint/Jint.csproj -c Release (all TFMs): ✅
• Jint.Tests net10.0 + net472: ✅
• Jint.Tests.PublicInterface net10.0 + net472: ✅
• Full Test262: ✅ (99,260 passed / 0 failed)

🤖 Generated with Claude Code