gb: Use only 4-LSB in LCD data

[deltabeard]

The LCD pixel data that Peanut-GB produces uses bits 0 and 1 for shade data, and bits 4 and 5 for layer data. If only bits 0-3 are used, then colour mapping using a look up table can be faster. A basic test on the RP2040 showed an improvement of approximately 8 FPS with this change.

This will require an API change.

[deltabeard]

The layer data is masked from the registers on the DMG. When this data is set by the ROM, Peanut-GB can store these values in a separate variable in bits 2 and 3, for use in a faster look-up table when rendering the LCD. Where bits 0 and 1 are for shade data, and bits 2 and 3 are for layer data. This should then allow for 4-bit colour in the LSB of a single byte. Further work would then be required to get packed 4-bit data.