Fix conda failure for onnxruntime-directml (microsoft#21526)

The change in microsoft#21005 works for directly building wheels with `build.py`,
but ort-nightly-directml wheels, as well as the 1.18.1 release of the
onnxruntime-directml python wheel, still do not work with conda since
they're built from the `py-win-gpu.yml` pipeline, which uses
`install_third_party_deps.ps1` to set compile flags.

Author: maggie1059

tools/ci_build/github/windows/install_third_party_deps.ps1

@@ -27,7 +27,7 @@ $Env:CMAKE_PREFIX_PATH = "$install_prefix"
 New-Item -Path "$install_prefix" -ItemType Directory -Force
 
 # Setup compile flags
-$compile_flags = @('/MP', '/guard:cf', '/DWIN32', '/D_WINDOWS', '/DWINVER=0x0A00', '/D_WIN32_WINNT=0x0A00', '/DNTDDI_VERSION=0x0A000000', '/W3')
+$compile_flags = @('/MP', '/guard:cf', '/DWIN32', '/D_WINDOWS', '/D_DISABLE_CONSTEXPR_MUTEX_CONSTRUCTOR', '/DWINVER=0x0A00', '/D_WIN32_WINNT=0x0A00', '/DNTDDI_VERSION=0x0A000000', '/W3')
 $linker_flags=@('/guard:cf')
 
 if ($use_cache) {