Jit_Integer: Use SHLX, SHRX, SARX

[merryhime]

Avoid unnecessary MOV instructions and RCX use when BMI2 is available.

[merryhime]

Implementing this exposed a bug where upper 32 bits of GPR registers were not zero (from existing JIT code it seems intentional for these bits to always be zero).

This bug exists in the existing implementation but is less commonly hit in the existing code because in order for it to be triggered it required a == s for a srwx instruction as well as a previous instruction sequence resulting in non-zero bits in the upper half of a GPR.

There are two ways to fix this:
(a) Never assume the upper half of a GPR is all zeros.
(b) Fix all code that violates the assumption that upper half of a GPR is all zeros.

After some debugging, the above only occurs in cases when Rs is in memory.

[merryhime]

Fixed.

In srwx, allowing Rs to reference memory had resulted in a 64-bit load rather than the desired 32-bit load, resulting in invalid data in the upper half of the register which was then shifted to the right.

<0.5% speed improvement in povray.

[merryhime]

@Sintendo Thanks, makes sense to me, applied.