style: comments carry the measurement, not the essay around it

[sotashimozono]

Measured before touching anything: 1479 comment lines over 6207 of code — 15% in src, 21% in
test. 102 blocks were four lines or longer, with 646 lines inside them.

The rule applied: keep the number, the version it was measured on, and the fact that is not
visible from the code. Drop the narration, the restatement of what the code does, and the
paragraph explaining why it matters — that last belongs in the commit that made the change, which
is where it already is.

	before	after
blocks of 6+ lines	23	12 (mostly file headers)
comment lines	1479	1259
src	15%	13%
test	21%	19%

Example:

# before — 6 lines
# The documented-and-marked definition is the shape this package asks for — `audit` wants
# a docstring on every public name AND a mark on the unfinished ones — and it was the one
# shape whose coverage came back `missing`, because no sibling started on `mk.line`. The
# span is still this sibling's; only where it starts moved.
(_is_doc_call(st) && a < line <= b) && return (line, b)

# after — the reason lives on `_is_doc_call`, two lines, once
(_is_doc_call(st) && a < line <= b) && return (line, b)

Nothing carrying a measurement was deleted. 1.03x / 0.985x / 3.76x, 0.601s vs 0.039s,
2/4 segmentation fault, 1916 candidates and 11026 methods, the three Julia versions the
world-counter claim was checked on, and the --depwarn=error inversion are all still there — in
fewer words.

1253 assertions, green; docs build clean.

🤖 Generated with Claude Code

[github-actions]

📚 Docs preview: https://codes.sota-shimozono.com/ExperimentalAPI.jl/previews/PR34/

(updates on each push to this PR)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!