[llvm] [Offload] Allow allocations to be freed without knowing their type (PR #140889)

[Callum Fare via llvm-commits]

================
@@ -639,6 +636,44 @@ struct CUDADeviceTy : public GenericDeviceTy {
     return OFFLOAD_SUCCESS;
   }
 
+  int free_non_blocking(void *TgtPtr) override {
----------------
callumfare wrote:

If the plan is to remove it eventually I think the current implementation is fine. Doing async allocs for everything sounds like it could introduce overhead from having to constantly synchronize the stream. 

https://github.com/llvm/llvm-project/pull/140889