[happykawayigt]

Description

Fixes #5270.
Closes #5475
Closes #5305

Long chains of single-token subscript trailers repeatedly rebuild the same prefix Line and BracketTracker state during recursive right-hand splitting. This patch detects that narrowly defined no-comment shape, reuses the already validated leaf metadata, caches cumulative ASCII widths for prefix heads, and falls back to the existing path for complex subscripts, comments, non-ASCII text, or other bracket shapes.

The formatting result is unchanged. In local CPython 3.12 benchmarks for x = a[0][1]... at the default line length:

Subscripts	Before	After
80	1.48 s	0.25 s
160	11.08 s	0.80 s
320	45.43 s	1.80 s

Output SHA-256 matched the base branch at every benchmark size. A 135-case matrix covering numeric, name, and string subscripts across multiple line lengths and contexts also matched byte-for-byte.

Validation:

• pytest tests/test_format.py -q (230 passed)
• pytest tests/test_black.py -q -k "not symlink" (160 passed, 6 deselected, 8 subtests passed)
• pre-commit on changed files: isort, flake8, mypy, EOF, trailing whitespace passed
• git diff --check
• New regression test bounds repeated BracketTracker.mark() calls and verifies idempotence

The three excluded test_black.py cases require Windows symlink privileges and fail locally with WinError 1314; they are unrelated to this change.

AI assistance was used during implementation. I independently reproduced the issue, profiled the split path, reviewed the final diff, compared outputs against an untouched base worktree, and ran the validation above.

Checklist - did you ...

• Implement any code style changes under the --preview style, following the stability policy? (No style change.)
• Add an entry in CHANGES.md if necessary?
• Add / update tests if necessary?
• Add new / update outdated documentation? (No documentation change is needed.)

[github-actions]

diff-shades results comparing this PR (5c1e307) to main (c7801d9):

--preview style: no changes

--stable style: no changes

---

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

[cobaltt7]

This exact number could get confusing to update if formatting changes.
Not sure that it even makes sense to check the formatting, the test is of the performance, not of the formatting.

[cobaltt7]

Could you add comments throughout this code? This is a fairly large code change and I find it difficult to follow everywhere - particularly the changes inside bracket_split_build_line().