Mark 67 code entries as copied rather than translated

[tamnd]

Built with pydocvi 0.1.8 (tamnd/python-docs-vi-translator#55 and #56).

The block rule now reads three more shapes of code, so 67 entries move out of prose:

def f(pos1, pos2, /, pos_or_kwd, *, kwd1, kwd2):
parrot(1000)                                          # 1 positional argument
parrot('a million', 'bereft of life', 'jump')

A signature quoted without its body, a line with a comment on the end, and a call whose arguments are quoted strings. All 67 were read one at a time before the rule shipped and there is no prose among them.

Five entries here had been sitting translated and not fuzzy with nothing behind them, because the segments that explained them left the memory when the classifier got stricter. S04 reported all five. They now carry # pydocvi: passthrough=literal_block. The text is unchanged, because it was already the source byte for byte, which is what L01 was reporting them for.

The rest of the diff is the header stamp and the writer's line wrapping on the entries it rewrote. No translated string changes in this PR.

Audit

check	before	after
S04 no unmarked machine translation	5	0
A01 the coverage report is current	4	0
L01 the translation is in Vietnamese	9	4
L02 no entry is the English verbatim	144	139
total findings	1,459	1,439
checks passing	28 of 41	30 of 41
hard checks failing	6	4

L01's remaining 4 are 3 copyright notices and a transcript of a Monty Python sketch. Both are policy questions about what should be translated at all, not defects.