Patch the decoding of fields containing ":" in the data

The split is done only once on each field; this has the effect of
splitting only with the first ":" and not with the ones that are present
as data.

The documentation is updated to reflect this new decoding of the files.