Fill solid optimizations per-color

[almindor]

This adds optimized (per color type) fill_solid paths as part of improving on issue #142.

This improved fill_solid fullscreen render speed on a ST7789 using 80Mhz SPI interface with esp32-c6 board running at 160Mhz CPU speed from ~48ms to ~18ms.

NOTE: the buffer size is arbitrarily chosen, bigger ones can get even better results, but I wanted to avoid using too much memory. If batch support is disabled these paths are not used. I hope we can come up with a better story for buffers overall in the future (e.g. wrt. to display-interface as well as this driver).

[almindor]

I'm also considering refactoring this using https://crates.io/crates/disjoint_impls/0.8.0 which should allow removal of the extra Model method and just doing it directly in the graphics module.

[almindor]

Closing in favour of #146 , however if for some reason we don't want to use the specialization "hack" we can fall back to this.