Fixed a bug in obs index inconsistency between matching and edge objects

[menelausyjl]

Obs index has been +1 when calling self._match_non_emitting_states_end(cur_lattice, obs_idx + 1, obs_next, lattice_best, expand=expand) I think here obs_idx is already the index for the next observation in this context, hence no need for the extra +1, to keep in accordance with the obs coordinates and edge object.