For each input sequence, the frequency of every word of length K is counted.
Counting is done with overlap. The counting itself is done by a C++ function.
With col.names=TRUE the K-mers are added as column names, but this makes the
computations slower.