Use 64-bit BYTE_MUL() for improved performance in blitting

By using godbolt.org compiler explorer it was found out that ARM Cortex-M4 (ARMv7-M) surprisingly supports 64-bit multiplication in 64-bit version of BYTE_MUL() and results in clearly shorter instruction sequence for critical code path in blitting.

Implement, test and document the 64-bit version in runtime.