Editorial: Improve internal and external consistency of RegExp pattern semantics

[gibson042]

No description provided.

[bakkot]

This is generally great, thank you. Had a few small comments.

[michaelficarra]

LGTM other than the 1-based indexing commit.

[bakkot]

@gibson042: are you OK with dropping e82e75b and 1f968cd from this PR and continuing discussion of those bits elsewhere? I know your preference is to land those as well before continuing the discussion, but @michaelficarra disagrees; the rest of this is uncontroversial and I don't want to block it on that point.

The other commits will also need rebasing into a collection of commits which can individually land; I'm happy to do that myself if you'd like.

[gibson042]

@bakkot Done; I extracted the controversial commits into #2150 so they aren't lost.

[bakkot]

Thanks! I rebased the review commits into the applicable commit, so that these commits can all land rather than being squashed.

[syg]

Generally lgtm, thanks.

As a follow-on I'd like to see us remove "mathematical sets", which AFAICT is only used for sets of characters in the RegExp machinery, and use the already defined Sets.

[syg]

For WordCharacters and CharSet, we could use the specification Set type, though the prose would need to be updated to say it is used in more places than just the memory model.