Q4 GEMM Improvement

CUrently in q4 gemm, a single block is being processed in KERNEL_mxn
functions. For KERNEL_8x8, 8 rows -> each row, 1 block is being
processed. I tried to modify this KERNEL_8x8 to process 8 rows -> 32
blocks once.

Signed-off-by: Shalini Salomi Bodapati <[email protected]>

Author: shalinib-ibm