[ci] Windows: Add appropriate CUDA flags

Add `--allow-unsupported-compiler` and
`ALLOW_COMPILER_AND_STL_VERSION_MISMATCH` flags to prevent fatal errors
induced by the version of the MSVC compiler, which has been updated
in the GitHub image to a recent version that by default does not support
any version of CUDA that is below 12.1.

.github/workflows/continuous-integration.yml

@@ -226,6 +226,8 @@ jobs:
                              -DALICEVISION_BUILD_STEREOPHOTOMETRY=OFF
                              -DALICEVISION_BUILD_SEGMENTATION=OFF
                              -DBOOST_NO_CXX11=ON
+                             -DCUDA_NVCC_FLAGS=--allow-unsupported-compiler;-D_ALLOW_COMPILER_AND_STL_VERSION_MISMATCH
+
           # This input tells run-cmake to consume the vcpkg.cmake toolchain file set by run-vcpkg.
           cmakeBuildType: Release
           buildWithCMake: true