Squaring requires fewer multiplications than a generic multiplication,
so take advantage of this when computing H^2 from H^1. This saves a few
instructions in gcm_init_vpclmulqdq_avx2 and gcm_init_vpclmulqdq_avx512.
Change-Id: I528bd877a53bb2fe5fa03067eda100a81192030f
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/78287
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>