Optimize ZUC-EEA3 for non-64 byte multiple buffers

When encrypting/decrypting buffers that are not 64-byte multiple long,
the remaining last bytes (less than 64 bytes) were calculated
by generating 64 bytes of keystream.

Instead, less keystream can be generated (in multiple of 4 bytes,
which is the size of a ZUC word), saving extra cycles.