refactor the GrayA44 pixel

[ripytide]

I'm not sure #138 should have been merged given the number of warnings and bugs in it unless I am misunderstanding the code. Regardless, I've attempted to fix the issues with this PR.

[ripytide]

@burbokop you might want to review these changes to see if they agree with your usecase.

[burbokop]

I used different formula. Instead of passing v and a to new as values in range 0..16 and doing (v << 4) | a my idea was to pass values in range 0..256 and compress its precision like this: ((v / 16) << 4) | (a / 16) which can be simplified as ((v >> 4) << 4) | (a >> 4) because v / 2^n = v >> n and then ((v >> 4) << 4) reduces to (v & 0xf0) and we have (v & 0xf0) | (a >> 4). But your version also works for me. I just need to change my code from new(v, a) to new(v / 16, a / 16) And the same logic with fn v() and fn a().