This PR fixes #100134

Problem lies in fact that regexp with u flag will change lastIndex to the start of surrogate pair if started after high pair:

const unicodeRe = /()/gu;
const regularRe = /()/g;
unicodeRe.lastIndex = 1;
regularRe.lastIndex = 1;
console.log('unicode', unicodeRe.lastIndex, unicodeRe.exec('💻').index, unicodeRe.lastIndex); // unicode 1 0 0
console.log('regular', regularRe.lastIndex, regularRe.exec('💻').index, regularRe.lastIndex); // regular 1 1 1

Which ended attempts to step forward for zero-length matches in endless loop.

This PR takes into account surrogate pairs and propagates lastIndex accordingly.
Endless loop during find matches is fixed in this PR.

Not in this PR:

1. In some cases (for files with a lot of matches) it's impossible to use Find next button to get to the next match after surrogate pair. Can give a try in this PR if needed, but would prefer to leave it out of scope of this PR
   (new fallback logic not triggered in that case, but should use similar strategy - remember last position and advance for zero-length matches, but this should survive reset)
   What do you think?

2. Inconsistency of handling matches of for surrogate pairs between JS RegExp's with u flag and ripgrep
   I can create issue with more details but simple example will be for file with 1💻 replace using RegExp 1. - Find in files breaks surrogate, while replace in editor replaces whole emoji