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

[Joseph Huber 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>,
----------------
jhuber6 wrote:

At least to me, what we're trying to build here is a generic offloading library that can handle all kinds of heterogenous setups. I like the HSA model for this (Which honestly is kind of close to this but never got its feel off the ground for anything non-AMD).
```
ol_memcpy(void *dst, ol_device_t dst_device, void *src, ol_device_t src_device, size);
ol_memcpy_async(ol_queue_t queue, void *dst, ol_device_t dst_device, void *src, ol_device_t src_device, size);
```
Would be my preference.

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