Keep import in English, at glossary v2

[tamnd]

Built with pydocvi 0.1.9 (tamnd/python-docs-vi-translator#57).

L02 no longer reports an entry whose whole msgid is a term the glossary keeps in English. The decision moves out of the check and into a written row that G03 then checks in both directions.

Most of it was already written. sys, builtins, exec and NaN were keep_en rows in v1 already, and what was missing was the check reading them. import is the one row this PR adds, for the 5 index entries that read import and name the statement.

Infinity is deliberately not added

The proposal had it. Measuring it first is what stopped it:

	entries
msgid is exactly Infinity	1
"infinity" inside prose, meaning the mathematical concept	80

vô cực is right for the 80 and wrong for the 1, and a row cannot say both. The prose wins, and that single index entry stays an L02 finding rather than 80 entries being told to un-translate a word.

Audit

check	before	after
L02 no entry is the English verbatim	139	86
G03 kept terms stayed in English	81	94
G06 no entry is on a stale glossary	0	860
total findings	1,439	2,259
checks passing	30 of 41	29 of 41
hard checks failing	4	4

G03 at 94 is the same change read the other way: 13 entries where import appears in prose and the translation rendered it in Vietnamese anyway. The corpus said both things, and now one of them is a finding.

G06 at 860 is the version bump itself. Every machine-translated entry was made against v1, stale --glossary is the command that re-queues them, and that needs a fleet run. It is a soft check and it is counting rather than failing, which is what its docstring says it is for.

Follow-up on the terminology this exposed: tamnd/python-docs-vi-translator#58. The glossary keeps list, type and Boolean in English and has no opinion about object or module, and all five are the same kind of Sphinx index category.