[KERNAL] PRIMM: Add 65C816 native mode support and optimize the code

[Fulgen301]

PRIMM is the only KERNAL library function left that uses abs,x addressing for stack access, thereby not working with a 65C816 in native mode where the stack page might have been moved. This PR adds an implementation that uses the 65C816's stack-relative addressing to work with arbitrary stack pointers; it also optimizes the 65C02's code path.

[Fulgen301]

With the new implementations,

jsr PRIMM
.asciiz "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"

now takes 36493 cycles instead of 38703 cycles on the 65C02 and 36490 cycles on the 65C816. (Measurements taken via the emulator's cycle counter).
The implementations essentially use the exact same code, with the 65C816 leveraging its stack addressing modes for lda $XX,S instead of lda $01XX,x.

[mooinglemur]

Thanks!