More correctly, *also* per-token. Because as it turns out, while the top-level edition affects parsing (I think), the per-token edition affects escaping of identifiers/keywords.
I expected this to be faster (due to less allocations and better cache locality), but benchmarked it is not (neither it is slower). Memory usage, however, drops by ~50mb (of `analysis-stats .`). I guess tt construction is just not hot.
This also simplifies using even less memory for token trees by compressing equal span, which I plan to do right after.
Some workflows are more easily expressed with a flat tt, while some are better expressed with a tree. With the right helpers, though (which was mostly a matter of trial and error), even the worst workflows become very easy indeed.
This PR touches a lot of parts. But the main changes are changing
`hir_expand::Name` to be raw edition-dependently and only when necessary
(unrelated to how the user originally wrote the identifier),
and changing `is_keyword()` and `is_raw_identifier()` to be edition-aware
(this was done in #17896, but the FIXMEs were fixed here).
It is possible that I missed some cases, but most IDE parts should properly
escape (or not escape) identifiers now.
The rules of thumb are:
- If we show the identifier to the user, its rawness should be determined
by the edition of the edited crate. This is nice for IDE features,
but really important for changes we insert to the source code.
- For tests, I chose `Edition::CURRENT` (so we only have to (maybe) update
tests when an edition becomes stable, to avoid churn).
- For debugging tools (helper methods and logs), I used `Edition::LATEST`.
