Sponge speed-up

This release utilizes System.arraycopy and avoids modulo computations in some of the inner loops inside the Sponge class. This provides a slight speed-up across all the configurations.

Fix unsigned remainder computation

Use Long.remainderUnsigned to compute remainders for 64 bit unsigned integers. This chooses the same columns in the memory matrix as the reference implementation does. Previously, when the number of columns was not a power of two, the remainder was sometimes computed differently.