Atomic instruction execution

[kremi151]

Currently instructions are split up into several operands, where one operand is executed at a time.
This was implemented to have emulation that's closer to how the real hardware executes code.
But this comes with some performance overhead, e.g. the 3DS and Android implementations suffer from slow emulation.
One way to mitigate this is to switch back to atomic instructions, i.e. executing directly an entire instruction at a time. When upating the internal system clock correctly, this won't cause any issue as the end result will be the same, with improved performance.