fix: Fix tokenizer hitting tail-recursion limit

[kitten]

Summary

TypeScript also has a tail recursion limit and the changes in #111, while being very effective at reduing parsing overhead times, caused us to now hit the limit of 1,000 when too many ignored tokens were being tokenized.

This PR addresses this by splitting the recursion of ignored tokens back out, reducing the iteration count of tokens for each chain of ignored tokens back to 1.

That does mean that we still have a limit of 1,000 tokens, including chains of ignored tokens, but from preliminary testing it looks like this is quite reasonable and hard to hit in an app with composed fragments.
I actually like that this may cause us to have a hard limit on which queries are reasonably parsed with gql.tada, and when hitting this limit, there's a high chance of a fragment boundary being missed.

Set of changes

• Readd skipIgnored type alias to tokenizer
• Restore old structure of takeSelectionRec type
  • Note: This is unrelated, but while it looked like it was better, it actually wasn't

[changeset-bot]

🦋 Changeset detected

Latest commit: 3a757fd

The changes in this PR will be included in the next version bump.

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[kitten]

Merging as a hotfix, so this isn't blocking anyone