CUDA: Fix thread flags for compilation and device linking
Since !1634 (merged) we consider every link item during device linking. However, items that start in
- may be host-specific link flags that nvcc will not understand during device linking. Filter such items using a white list. In particular, this allows
-pthread to be used for host linking while not polluting the device link line.
Also teach FindThreads to wrap
-Xcompiler for compilation through nvcc.
Update the test suite to cover use of pthreads with CUDA.
Fixes: #18008 (closed)