Generate better assembler for modulation 6 bits as the previous version's code generation was considerably suboptimal

The compiler was not able to generate vectorized code for the modulation of 6 bits code e.g., https://godbolt.org/z/a4ncf4Krf So I vectorized the function 12 bits to 16 bits. In my computer is roughly 4-5x faster e.g., 445 ns vs 1818 ns in the test file provided at godbolt link and I think that the code is better now as the intention of the programmer is clear using a function i.e., unpack12to16 rather than code without a name.