Delay calculate flags

[fengb]

It seems like in general, CPU computation flags are not used very often. It would be nice if we could restructure the code to delay flags until they're required.

New Op layout:

const daa = struct {
    fn exec(cpu: *Cpu, mmu: *Mmu, arg0: Arg, arg1: Arg) Result(1, .{4}) {}

    fn flags(val: u8, prev: u8) Flags {}
    // or
    fn flags(val: u8, prev: u8, existing: Flags) Flags {}
    // or none
};