perf(graphics): implement faster fill_solid

[techmccat]

Added a fill_solid implementation to draw large chunks much faster.
Managed to get my use-case (drawing quadtrees) from 40ish ms to less than 10ms.

I did a bit of testing and it seems to be working as intended when rotated too.

[jamwaffles]

Thanks for the PR! I'll drag my test rig out tonight and test/review on my end.

[jamwaffles]

First, thanks for taking the time to implement this PR, however I saw the coded and thought I could make a slightly cleaner implementation in #28 (and an animated example with RTIC).

I didn't check yours before I wrote my impl, but I think the impl in this PR would run into issues with rectangles smaller than 8 bits high, as well as the top/bottom partial blocks - I believe they would both clobber any existing data in the current byte as they don't do any kind of bitwise OR-ing with the current value.

Sorry to step on your toes with this, but would you be ok closing this PR in favour of #28?

[techmccat]

I'm not overwriting anything here though, partial blocks just fall back to the old pixel by pixel drawing.
Your implementation is better in this aspect, and not having recursion is a lot better too.

[jamwaffles]

I'm not overwriting anything here though, partial blocks just fall back to the old pixel by pixel drawing.

Right, apologies for stating otherwise - I forgot about the recursion when I wrote my comment.

Your implementation is better in this aspect, and not having recursion is a lot better too.

Yeah, definitely nice to steer away from recursing on these small MCUs :)