Release notes with the version in the filename are history too

[melbinjp]

release[-_ ]?notes has been on _HISTORICAL_NAMES since the fastapi measurement, where a single release-notes.md produced 162 findings, more than every other document in twenty repositories combined.

It could not see its own convention when a version was stamped on the filename. The pattern required an extension or a slash immediately after the name:

(^|/) (?: NAMES ) (?: \.[a-z0-9]+ )? (?: / | $ )

docs/release-notes.md matches. docs/release_notes_v7.9.0.md does not, because _v7.9.0 is neither.

Where it showed

enarjord/passivbot, sweep batch 10. Two of its five findings:

document	says	truth
docs/release_notes_v7.9.0.md:21	configs/examples/default_trailing_grid_long_npos10.json	deleted in 565376e9c, four months after v7.9.0
docs/release_notes_v7.10.0.md:8	configs/examples/default_trailing_grid_long_npos7.json	deleted in 71efc3c57

Both documents are correct. The config existed at the version the filename is named for, and nobody rewrites the v7.9.0 notes to track a tree they were never describing. This is the changelog argument exactly, applied to a project that keeps one file per release instead of one long file.

Pillow escapes it already, for a reason that has nothing to do with the filename: its notes live in docs/releasenotes/2.3.2.rst, and the DIRECTORY rule reads that. A project that keeps them flat got judged for identical content.

What changed, and how far

(^|/) (?: NAMES ) (?: [-_ ] v? \d+ (?: \.\d+ )* )? (?: \.[a-z0-9]+ )? (?: / | $ )

The version has to follow immediately and has to be a number. changelog-policy.md, release-notes-process.md and history-of-the-parser.md stay judged, and there are assertions on each.

Measured across every sweep batch, 147 finding lines in total: this suppresses exactly the two above and nothing else. passivbot goes from 5 broken to 3, and the three that remain are real.

197 tests.