Move call to MLAS_CPUIDINFO::GetCPUIDInfo() out of MlasSQNBitGemmDispatchNeon initialization.

[edgchen1]

Description

Move call to MLAS_CPUIDINFO::GetCPUIDInfo() out of MlasSQNBitGemmDispatchNeon initialization.

Motivation and Context

Reduce binary size when MatMulNBits op is not included in the build.

I believe the side effect of MLAS_CPUIDINFO::GetCPUIDInfo() (e.g., initializing a static object) prevents the linker from discarding the code in a build where the associated MLAS functions are unused.

bloaty diff output of Android minimal build with no ops:

baseline: 06482c2, updated: 8f5d213

$ bloaty -d compileunits ${UPDATED_DIR}/libonnxruntime.stripped.so --debug-file=${UPDATED_DIR}/libonnxruntime.so -- ${BASELINE_DIR}/libonnxruntime.stripped.so --debug-file=${BASELINE_DIR}/libonnxruntime.so
    FILE SIZE        VM SIZE
 --------------  --------------
  +0.1%     +85  +0.1%     +85    [section .rodata]
   +12%     +60   +12%     +60    D:/source/onnxruntime_2/build/android.minimal_baseline/MinSizeRel/_deps/pytorch_cpuinfo-src/src/linux/multiline.c
  +3.5%     +44  +3.5%     +44    D:/source/onnxruntime_2/build/android.minimal_baseline/MinSizeRel/_deps/pytorch_cpuinfo-src/src/arm/linux/aarch64-isa.c
  -0.8%     -24  -0.8%     -24    [section .dynsym]
 -33.3%     -32 -33.3%     -32    [section .gnu.version_r]
  -0.1%     -53  -0.1%     -53    D:/source/onnxruntime_2/onnxruntime/core/session/onnxruntime_c_api.cc
 -100.1%     -96 -100.1%     -88    [11 Others]
  -4.9%     -96  -4.9%     -96    [section .got]
  -3.8%    -104  -3.8%    -104    D:/source/onnxruntime_2/build/android.minimal_baseline/MinSizeRel/_deps/protobuf-src/src/google/protobuf/arena.cc
  [DEL]    -224  [DEL]    -224    D:/source/onnxruntime_2/onnxruntime/core/mlas/lib/threading.cpp
  -0.6%    -244  -0.6%    -244    [section .eh_frame]
  [ = ]       0 -65.4%    -257    [section .tbss]
  -1.4%    -360  -1.4%    -360    [section .data.rel.ro]
  [ = ]       0 -25.8%    -456    [section .relro_padding]
  [DEL]    -876  [DEL]    -988    D:/source/onnxruntime_2/onnxruntime/core/mlas/lib/qnbitgemm_kernel_neon.cpp
  -1.6% -1.08Ki  -1.6% -1.08Ki    [section .rela.dyn]
  -0.8% -1.71Ki  -0.8% -1.71Ki    [section .text]
 -15.6% -2.62Ki -15.6% -2.62Ki    D:/source/onnxruntime_2/onnxruntime/core/common/threadpool.cc
  [DEL] -6.97Ki  [DEL] -6.97Ki    D:/source/onnxruntime_2/onnxruntime/core/mlas/lib/sqnbitgemm_kernel_neon_fp32.cpp
  [DEL] -7.18Ki  [DEL] -7.18Ki    D:/source/onnxruntime_2/onnxruntime/core/mlas/lib/hqnbitgemm_kernel_neon_fp16.cpp
  [DEL] -10.1Ki  [DEL] -10.1Ki    D:/source/onnxruntime_2/onnxruntime/core/mlas/lib/sqnbitgemm_kernel_neon_int8.cpp
  -2.6% -31.6Ki  -2.6% -32.4Ki    TOTAL

[edgchen1]

binary size checks pipeline runs:
baseline:
https://dev.azure.com/onnxruntime/onnxruntime/_build/results?buildId=1644879&view=logs&j=31d11bb5-e65f-5534-44ce-dcac1fb18dd3&t=51484273-4d20-5e69-1970-1124d6374185&l=3072
updated:
https://dev.azure.com/onnxruntime/onnxruntime/_build/results?buildId=1645858&view=logs&j=31d11bb5-e65f-5534-44ce-dcac1fb18dd3&t=51484273-4d20-5e69-1970-1124d6374185&l=3072