only makes sense, and seem to covers all use cases for palettes.

is it better? is it worse? dunno...
feel like vector should be more permissive with conversion now, for
consistency, but too afraid to tackle that now.

that was inherited from geom::vector.

is it a thing here or am I just working around SDL pretending it to be a
thing here? Cause I tried just rgba8888 on my other little endian
machine and it worked perfectly, and on current one it doesn't, but the
display on other one broke so I can't test if this stuff works there -_-

not too bad...

wrappers with blend mode parameter, and a pixel view version.

and of course it'd only break after a while.

Interpolated getter analogous to the setter and a separate, dimension agnostic bound checking for both.

mainly splitting it into .h,.hpp,.cpp, providing template instantiation
for all available pixel formats (4 now, more in the future).
Also some small improvements and fixes that I wasn't able to keep track
of because of the restructure -_-
Why can't I do things one at a time? >.<

for when SDL doesn't provide those.

Also rgb888 is actually xrgb8888??? weird... will need to provided
actual rgb888

and a couple of minor const noexcept changes.

as in N^2 long...