add some semantic function of instructions

[yjxxin]

When I added semantic functions, I found that there are many instructions in remill that did not add corresponding semantic functions, such as VPMOVSXBQ and VPMOVSXWD. I added more than 300 similar instructions when doing related work, such as

namespace {
template <typename D, typename K, typename S>
DEF_SEM(VPMOVSXBQ_MASKmskw_SIMD128, D dst, K k1, S src) {
  auto src_vec = SReadV8(src);
  auto dst_vec = SClearV64(SReadV64(dst));
  auto k_vec = Read(k1);
  for (int i = 0; i< 2; i++) {
    if (READBIT(k_vec, i) == 0) {
      dst_vec = SInsertV64(dst_vec, i, 0);
    }else {
      auto v = SExtTo<int64_t>(SExtractV8(src_vec, i));
      dst_vec = SInsertV64(dst_vec, i, v);
    }
  }
  SWriteV64(dst, dst_vec);
  return memory;
}
}

#if HAS_FEATURE_AVX512
DEF_ISEL(VPMOVSXBQ_XMMi64_MASKmskw_MEMi8_AVX512) = VPMOVSXBQ_MASKmskw_SIMD128<VV128W, R8, MV16>;
DEF_ISEL(VPMOVSXBQ_XMMi64_MASKmskw_XMMi8_AVX512) = VPMOVSXBQ_MASKmskw_SIMD128<VV128W, R8, V128>;
#endif  // HAS_FEATURE_AVX512

namespace {
template <typename D, typename K, typename S>
DEF_SEM(VPMOVSXWD_MASKmskw_SIMD128, D dst, K k1, S src) {
  auto src_vec = SReadV16(src);
  auto dst_vec = SClearV32(SReadV32(dst));
  auto k_vec = Read(k1);
  for (int i = 0; i< 4; i++) {
    if (READBIT(k_vec, i) == 0) {
      dst_vec = SInsertV32(dst_vec, i, 0);
    }else {
      auto v = SExtTo<int32_t>(SExtractV16(src_vec, i));
      dst_vec = SInsertV32(dst_vec, i, v);
    }
  }
  SWriteV32(dst, dst_vec);
  return memory;
}
}

#if HAS_FEATURE_AVX512
DEF_ISEL(VPMOVSXWD_XMMi32_MASKmskw_MEMi16_AVX512) = VPMOVSXWD_MASKmskw_SIMD128<VV128W, R8, MV64>;
DEF_ISEL(VPMOVSXWD_XMMi32_MASKmskw_XMMi16_AVX512) = VPMOVSXWD_MASKmskw_SIMD128<VV128W, R8, V128>;
#endif  // HAS_FEATURE_AVX512

If you are satisfied, I can upload all the instructions I have added to the community to help remill expand and support conversion of more instructions.

[pgoodman]

Mostly superficial changes requested. We'd be very interested in those other 296 semantics!

[pgoodman]

It looks like this relies on the K registers to be in place, so I'll wait on that PR to be merged. Otherwise, excellent work!