Re-apply under pydocvi 0.1.4

[tamnd]

tamnd/python-docs-vi-translator#47 narrowed the identifier rule. is_version_marker used to call any single token of identifier characters an identifier, and a single English word is a single token of identifier characters, so apply copied 2 808 headings through from the msgid and stamped each one passthrough=version_marker. That stamp is a claim that the string needs no translation.

Written back as English	Entries
Availability	62
Exceptions	25
Author	21
Introduction	18
Description	15
Notes	11

This run drops those msgstrs, which puts the entries back to untranslated, which is what they are.

What the numbers do

	Before	After
Passthrough	13,900	11,074
Untranslated	70,934	73,760
Tier 1 untranslated	2	31

None of that is new breakage. The strings were English on the page before this and they are English on the page after it. What changed is that the corpus no longer claims they were meant to be.

apply --check is clean and S08 is back to 0, so the blocking gate holds.

Why the audit total goes up

1,270 findings to 1,410, for the same reason in reverse. Narrowing the rule puts 6,558 human-written entries into Corpus.prose() for the first time, so the checks that read a translation can finally see them.

L02 goes from 10 findings to 144 over 20 distinct msgids. About 96 are ordinary words a reviewer left in English: module 41 times, object 36, type 12, Boolean 4, Lists 3. The other 38 are identifiers a reviewer was right to keep, sys 20 and builtins 8 and import 5, and L02 has no way to tell those apart from the others yet. That wants a kept glossary entry per identifier, filed separately.

Still nine hard checks failing and the audit step stays advisory.