Create character sets for 1, 2 and 3rd positions

[gaurav]

We need to "hide" them in the output so we don't see them during import and complain of overlapping datasets. Possibilities:

NEXUS:

1. Hide it in the "ASSUMPTIONS" block, whence we can't read them.

TNT:

1. Specially formatted sets, in the format qr/.*_pos[0123]$/. Note that we can probably fiddle with TNT import so that we no longer export a single 'pos1' set, but multiple 'coi_pos1', 'coii_pos1', etc. sets and then reassemble them during import.

BOTH:

1. Use specially formatted names and ignore them on import.
2. Hide them in comments, so they can be uncommented by the users when they need to be used.

[gaurav]

Do this for TNT first, as that's more important.

[gaurav]

Done as of 9fa9cc9034, I hope. Testing is underway and is showing some suspicious results.

[gaurav]

Okay, the "suspicious" results mentioned above include duplication in the '_posX' sets.