Inside a BED window (i.e. not at the very edges), the total number of incoming transition counts to an M state should match the total number of outgoing transition counts in the next column, and the same for N states. Proposal is to add a test that verifies this