Three lang/ msgids never resolve, and nothing checks the msgid pairing

[xroche]

lang/*.txt are consecutive line pairs, an English msgid then its translation, and nothing validates that shape. Finnish.txt had ended with a blank line since 2012, which was harmless while it was the last line. #588 appended new strings after it, so the blank became interior and knocked every pair past it out of phase: all 14 strings #588 added to Finnish, one of which 938a873 later reworded into the CONNECT proxy string, were keyed off a translation instead of an msgid and rendered as English. Dropping the blank restores the phase. Italiano and Portugues-Brasil separately each carry one msgid that matches nothing in English.txt (ISO 9660 for ISO9660, Relative URL for Relative URI), so those two entries never resolved either.

The test matters more than the three fixes. It rebuilds the msgid set from English.txt and checks every file against it byte-wise, since each file is in its own declared legacy charset and lang.def is CRLF. It also checks lang.def's English values, which are the join key into the msgids, so rewording one without moving the msgids no longer silently drops all 30 translations. Both parses assert a floor on what they extracted, so a parse that breaks fails instead of passing on an empty set. On unfixed master it reports all three defects with line numbers; it passes here.

It does not check translation text, empty translations, or a file missing trailing entries, since a partial translation is legitimate and common. It also does not check the declared LANGUAGE_CHARSET: Latin-1 and Latin-2 decode any byte sequence, so a decode check would pass vacuously and could never catch the known-wrong Svenska header.

Context for #306 (Turkish), which needs the msgid join to be sound, but does not close it.