An RP2040 fix and a couple of parallel related changes

[Daft-Freak]

Two small fixes and a cleanup/optimisation found while working on parallel support for RP2040. (I actually have two versions of that, need to do some tidying...)

• Fixes a missing check from my previous PR, breaking writePixels for !HARD_SPI on RP2040.
• Fixes spiRead on AVR leaving RD low due to the TFT_RD_HIGH being in the #else branch (only noticed because I had another special case there at one point)
• Always uses the fast path for writeColor on a 16-bit bus, simplifying the else as it only has to deal with 8-bit now.