cxgb4: speed up on-chip memory read

Use readq() (via t4_read_reg64()) to read 64-bits at a time.
Read residual in 32-bit multiples.

Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>