Recent Solver Crashes

[travisgoodspeed]

Recent expansions to the solver have introduced regressions when the sizes didn't properly line up with implicit assumptions. We'll use this ticket to track their repairs, then close the ticket only when we have test cases that randomize sizes as well as contents.

[travisgoodspeed]

Reviewing a number of changes here:

1. Raw access to the bit tables has been removed.
2. The CLI will now crash on an illegal bit access; GUI will print a warning. (strictmode)
3. TLCS47 ROM is now honest about its hardcoded size. GatoROM TLCS47 Format #72 will track making it generic.
4. Unmatched rows now become empty ROMs.

I'm no longer able to crash the decoder or solver, so I'll resolve this issue. Any future problems should be manageable in the test corpus.