Removing pandas from countMismatches in EstimateError.py

[Roy Jiang]

cProfile suggests that indexing and slice operations related to using pandas dfs are slowing down countMismatches.

Removing pandas from estimateerror with either csv or numpy replacement.

[Jason Vander Heiden]

Also see Sequence.calculateSetError.

[Roy Jiang]

EstimateError.py still needs unit testing overall. Should be a noticeable performance difference with new branch EstimateErrorNoPandas.