fix(match): include last word in capture groups spanning wildcards

[xianjianlf2]

Fixes #1139

Symptom

A named capture group that wraps a greedy ./.* wildcard together with surrounding tokens drops the tokens after (and mis-sizes the span around) the wildcard:

nlp('one two three after').match('[<all>one . after]').group('all').text()
// actual:   'one two three'
// expected: 'one two three after'

Reproduced by @spencermountain in the issue thread.

Root cause

In src/1-one/match/methods/match/steps/astrix.js, doAstrix sets the group result with:

g.length = skipto - state.t

The assignment overwrites whatever length earlier regs in the same group had already contributed, so the group is re-anchored to just the wildcard span. The tokens captured before the wildcard are dropped from the group, and the trailing literal ends up outside the reported range.

Fix

Accumulate instead of overwrite:

g.length += skipto - state.t

Leading tokens stay in the group, and the trailing literal is appended by the normal match loop as before.

Tests

Added tests/two/match/greedy-capture.test.js covering the issue's repro plus variants (leading literal + wildcard, wildcard + trailing literal, group starting at the wildcard, and greedy min/max behavior). Full suite passes: 13,419 assertions, 0 failures (npm test).

[spencermountain]

wow, thank you!
this is very helpful!!