...that a more picky idxdict now complains about when it is not able to read an
integer count where it expects to find one. In most cases, a previous line
count was apparently wrong (similar to df423c53829faa96f46acd71887fd0f8697c49e9
"fix Lithuanian thesaurus"). But:
* ro/th_ro_RO_v2.dat had three cases that look like lines got oddly duplicated.
And the excessive whitespace in the middle of the line in the second case
looks odd, too.
* uk_UA/th_uk_UA.dat leaves me completely clueless. I edited it so that it
passes through idxdict, but it definitely needs inspection from a native
speaker and/or expert in that thesaurus .dat format.