Implement forced 16 bit screenmodes

[ali1234]

You can set ScreenMode::lores_565 or ScreenMode::hires_565 to be
guaranteed a 16 bit framebuffer on every platform. On Pico, this
does exactly the same as the normal screenmodes.

[ali1234]

Why: 565 screenmodes have better performance because they are word aligned. The whole pixel can be read or written in a single memory access.

This is a bit hacky though. For some reason the RGB565 mode is BGR in SDL. On Blit it is RGB.

I have converted a few if/else blocks into switch statements. And I duplicated the flip functions in the 32blit HAL. I assume those are supposed to be as fast as possible.

[ali1234]

Also most of these files have a mixture of tabs and spaces, so I have not attempted to format the new code to match.

[ali1234]

Also I haven't tested the lores modes.

[Daft-Freak]

Also most of these files have a mixture of tabs and spaces, so I have not attempted to format the new code to match.

I think most of the 32blit code is two spaces (and some of the code is definitely going from 2 -> 4 here)

Looks like there's a R/B swap bit for DMA2D if that helps with getting it consistent (DMA2D_OPFCCR_RBS)

[Daft-Freak]

Okay yeah, I see what you mean about it being RGB on the blit. I think the blit display is getting R/B swapped somewhere because the LTDC RGB888 mode is actually BGR... (Thought it was st7272a_set_bgr, but that's never called).

The important thing is that this all seems to work though.

[ali1234]

Yes, on blit I was surprised that it "just works" where as on SDL I had to tell it the buffer is BGR, even though on SDL the 24 bit buffer is declared as RGB.

[ali1234]

I need to make a test program before I'll really trust this...

[ali1234]

Replaced by #730