In some NIST files, the field 999 containing binary data could be empty.
In this case, we shall set the value to None and continue the parsing of
the NIST file.

This is done because of a US NIST file that was off by one byte at the
end of the file. This can append if the last byte is not a "1C". If this
skip produce a valid NIST object (i.e there is not fields after), it's
probable that the entire NIST file has been processed correctly.

If this error is present in the middle of the file, the processing of
the next record should fail and crash completely the loading of the file
(hopefully...).

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.

This allows to check every fields in the load and dump process.

This file format abbreviation was found in a US NIST files. The field
contained a JPEG2000 file. I can only suppose that the format was LOSSY,
but I can not confirm nor disprove it. The file will be assumed to be a
JPEG2000 LOSSY.