Improve lifting of BEXTR instruction for x86

[SlidyBat]

What is the feature you'd like to have?
Lifting of the BEXTR instruction is currently handled as an intrinsic.

This produces hard to understand HLIL, and could be improved if it was lifted to a bit-shift/mask instead.

Assembly:

00000000  b902010000         mov     ecx, 0x102
00000005  c4e270f7c7         bextr   eax, edi, ecx
0000000a  c3                 retn

HLIL:

00000000    uint64_t sub_0(int32_t arg1)

00000005        uint64_t rflags
00000005        int32_t temp0
00000005        temp0, rflags = __bextr_gpr32d_gpr32d_gpr32d(arg1, 0x102)
0000000a        return zx.q(temp0)

IDA output:

__int64 __fastcall sub_0(unsigned int a1)
{
  return (a1 >> 2) & 1;
}

Is your feature request related to a problem?
Can replicate above assembly by pasting these bytes:
b902010000c4e270f7c7c3

Are any alternative solutions acceptable?
Separating the start/len of the bextr intrinsic into 2 separate arguments (and making intrinsic name shorter) could make it more readable, but bit-shift/mask would be best.

[xusheng6]

This is a very reasonable feature request! However, we are unlikely to deal with it very soon since the x86 instrinsics are all generated from xed instead of hand-rolled