CUDA/C++14

C++17 builds with NVCC are currently (up to CUDA 11.2)
failing with an internal compiler transformation bug
in extended lambda support.

.github/workflows/windows.yml

@@ -42,7 +42,7 @@ jobs:
 
   # Build libamrex and all tutorials with CUDA
   tutorials-cuda:
-    name: MSVC C++17 w/ [email protected] w/o Fortran w/o MPI
+    name: MSVC C++14 w/ [email protected] w/o Fortran w/o MPI
     runs-on: windows-2019
     steps:
     - uses: actions/checkout@v2
@@ -67,8 +67,8 @@ jobs:
               -DAMReX_FORTRAN=OFF        `
               -DAMReX_MPI=OFF            `
               -DAMReX_GPU_BACKEND=CUDA   `
-              -DCMAKE_CXX_STANDARD=17    `
-              -DCMAKE_CUDA_STANDARD=17   `
+              -DCMAKE_CXX_STANDARD=14    `
+              -DCMAKE_CUDA_STANDARD=14   `
               -DAMReX_CUDA_ARCH=6.0      `
               -DCMAKE_CUDA_COMPILER="C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v11.1/bin/nvcc.exe"
         cmake --build build -j 2