[llvm] [Offload] Implement the remaining initial Offload API (PR #122106)

[Callum Fare via llvm-commits]

================
@@ -16,8 +16,8 @@ def : Function {
     let details = [];
     let params = [
         Param<"ol_queue_handle_t", "Queue", "handle of the queue", PARAM_IN>,
----------------
callumfare wrote:

That's all fair, I've added an implementation of a generic async memcpy, along with a function to retrieve the host device. Does it look like what you had in mind?

I've kept the old ones around for comparison, although I've renamed them to memcpy variants for consistency.

I also agree it would be nice to have host task functionality like `cuLaunchHostFunc`.

https://github.com/llvm/llvm-project/pull/122106