feat: declare the four young layers @experimental, and let @entered be told where to write

[sotashimozono]

The package had 82 public names and declared 9 — the release layer, because its file format was a
guess. The four layers that moved this week were not declared. That is exactly the state this
package exists to make visible in somebody else's code.

What is now declared, and the measured reason each carries

layer	why it is not settled
record, Record, Hit, write_record, … (13 names)	Record gained a value field in #26, after write_record already had a file format — a file written by an earlier version reads back with that field empty. And paths+timing are refused together because the pair segfaulted 2 runs in 4 where each alone crashed 0 in 4.
@entered	text report, no schema, changed twice in its first week — sort order and column alignment (#26), singular verb in the footer.
reach, Reach, verdict, … (9 names)	the :clean/:unknown line is drawn by Julia's internal IR accessors, which differ by minor: Base.IRShow.getdebugidx on 1.12, codelocs/linetable on 1.11, Core.TypeEgal{T} where earlier versions say Type{T}.
verification, coverage, flush_coverage, … (7 names)	counts come from ccall(:jl_write_coverage_data, …) and the .cov line format. Julia documents neither.

39 of 82 declared, 43 not.

Why before a release rather than after

The package's own isbreaking is the arbiter. A declared name can change without the change being
breaking by this package's own definition; an undeclared one cannot. Record gaining a field one
commit ago is not the behaviour of a settled struct, and shipping it undeclared would lock it.

The test that used to say the opposite

test/test_dogfood.jl pinned the declared set to exactly the release layer's nine names and
asserted every reason contained "schema" — i.e. it encoded "the release layer is the only young
one", which this PR falsifies. It now carries a per-layer table whose anchor is a measured
detail ("segfault", "getdebugidx", "jl_write_coverage_data"), so a reason that could have
been written without doing the measurement fails.

An equality against a hand-written set has one failure mode: mark every name, update the set,
green. So a second testset asserts the 24 names of the settled core — @experimental, entered,
audit, mark, surface, … — are on the surface and are not marked. A package that declared
everything would be telling you nothing.

Also: @entered io expr

The report was hardwired to stdout, so a caller who wanted it in a log had to redirect the whole
process to get at one line. io resolves when the block runs, not when the macro expands, so the
default still follows redirect_stdout; the test's control asserts naming an io moves the
report rather than copying it.

1192 assertions, green.

🤖 Generated with Claude Code

[github-actions]

📚 Docs preview: https://codes.sota-shimozono.com/ExperimentalAPI.jl/previews/PR27/

(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!