Let one glossary row answer both questions about a word

[tamnd]

Closes #58.

L02 exempts an entry whose whole msgid is a glossary term kept in English, and #57 wired that to keep_en. That made one field carry two decisions: how the word reads inside a sentence, and whether an entry that is only the word is already correct.

For most rows those have the same answer. For six of them they do not.

term	on its own	in a sentence
float	a cell in the struct format-code table, next to _Float16	số thực
type	the argparse keyword, in a table of argparse keywords	kiểu
list	a cell in the json conversion table, next to array	danh sách
import	the statement, in the index of statements	nhập
slice	a glossary headword	cắt
callable	a glossary headword	gọi được

While the two questions shared a field a row could only answer one of them, and the answer that got written was whichever silenced the louder check. 69 of G03's 94 findings are correct translations of type and list, held there by rows that say "keep this in English" when what they meant was "leave the table cell alone".

So Term gains identifier, Glossary gains standalone, and L02 reads the second one. A keep_en row is standalone without being told, because a term that is English in every sentence is English on its own too. Nothing in the current file changes meaning and the new decisions are all made in the glossary rather than here.

The prompt says both things on one line. library/struct.po has float as a table cell two entries away from a sentence about floats, so a batch holds both readings and a line carrying only the rendering would get the cell translated.

- float -> số thực  (a string that is only this word names the thing: leave it in English)

The measurement is in the content repo's proposal and the rows are the second half of this. Alone, this commit moves nothing: standalone equals kept until a row sets the flag.

Local gate green: 1 422 tests, 97.46% coverage, mypy --strict clean, H03 clean. Not waiting on Actions.