BitUtils: cleanup constexpr usage for msvc clz

[shuffle2]

• Can use _BitScanReverse{64} for all archs we support with msvc
• Can actually support being used in constexpr evaluation contexts

see https://developercommunity.visualstudio.com/idea/434712/make-intrinsic-functions-supported-in-a-constexpr.html

here is a small test: compiling the following with cl /Ox /std:c++latest clz.cpp /link /debug

int main() {
    const uint32_t val32 = 9;
    const uint64_t val64 = 9;
    int clz1 = CountLeadingZeros(val32);
    int clz2 = CountLeadingZeros(val64);
    constexpr int clz3 = CountLeadingZeros(val32);
    constexpr int clz4 = CountLeadingZeros(val64);
    printf("%d %d %d %d\n", clz1,  clz2,  clz3,  clz4);
    return 0;
}

arm64:

int __cdecl main(int argc, const char **argv, const char **envp)
{
  unsigned int v3; // w10
  __int64 v4; // x1
  unsigned int v5; // w10
  __int64 v6; // x2

  v3 = __clz(9u);
  if ( 1 == (unsigned __int8)(v3 >> 5) )
    v4 = 32i64;
  else
    v4 = v3;
  v5 = __clz(9ui64);
  if ( 1 == (unsigned __int8)(v5 >> 6) )
    v6 = 64i64;
  else
    v6 = v5;
  sub_140003F94("%d %d %d %d\n", v4, v6, 28i64, 60i64);
  return 0;
}

x64:

int __cdecl main(int argc, const char **argv, const char **envp)
{
  unsigned __int64 v3; // rax
  __int64 v4; // rdx
  int v6; // [rsp+20h] [rbp-18h]

  v6 = 60;
  _BitScanReverse((unsigned int *)&v3, 9u);
  v4 = (unsigned int)(31 - v3);
  _BitScanReverse64(&v3, 9ui64);
  j_printf("%d %d %d %d\n", v4, (unsigned int)(63 - v3), 28i64, v6);
  return 0;
}