fix quadratic backtracking in unicode escape normalization

[sahvx655-wq]

normalize_unicode_escape_sequences runs UNICODE_ESCAPE_RE over every string leaf, and the pattern matched a backslash run with (\+) immediately followed by a required escape body. A long run of backslashes with no trailing u/U/x/N escape makes the engine retry the run from each restart position, so formatting a string of backslashes is quadratic: a 64KB literal already takes minutes, and blackd accepts far larger bodies over the wire. Making the escape body optional lets the whole run be consumed in a single match, which restores linear time and leaves the output identical.

[github-actions]

diff-shades results comparing this PR (b06aa63) to main (d246367):

--preview style: no changes

--stable style: no changes

---

What is this? | Workflow run | diff-shades documentation

[JelleZijlstra]

Thanks, could you add a changelog entry? Something like "Improve performance on strings containing many consecutive backslashes"

[sahvx655-wq]

Added it under Performance, thanks.