No consistency check on thermo group adjlists

[nyee]

When trying to export the RMG-Java database to RMG-Py, I found some errors in thermo groups that should cause RMG to crash. For example, one of the bond orders is 'H' in 3a0405a

4 * Cs 1 {1,S} {5,S} {9,H}

in line 81 of Radical_Dictionary.txt

I will commit corrections for the library, but there should probably be a check on RMG for valid adjacency lists here.