Enable CUDA device code warnings as errors (#307)

Adds `-Werror=all-warnings` NVCC flag to ensure all CUDA device code warnings are treated as errors. Only enabled on CUDA 11.2+ because CUDA 11.0 has PTXAS warnings that go away in newer CUDA versions.

Missed this in #299.

Authors:
  - Mark Harris (https://github.com/harrism)

Approvers:
  - Dante Gama Dessavre (https://github.com/dantegd)

URL: #307

cpp/cmake/modules/ConfigureCUDA.cmake

@@ -26,7 +26,9 @@ endif()
 list(APPEND RAFT_CUDA_FLAGS --expt-extended-lambda --expt-relaxed-constexpr)
 
 # set warnings as errors
-list(APPEND RAFT_CUDA_FLAGS -Werror=cross-execution-space-call)
+if(CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER_EQUAL 11.2.0)
+    list(APPEND RAFT_CUDA_FLAGS -Werror=all-warnings)
+endif()
 list(APPEND RAFT_CUDA_FLAGS -Xcompiler=-Wall,-Werror,-Wno-error=deprecated-declarations)
 
 # Option to enable line info in CUDA device compilation to allow introspection when profiling / memchecking