[core][rdt] Move copying of receiver side pool outputs onto its own cuda stream

[Sparks0219]

Description

Following up on #64487 we should move the copying of the receiver side pool outputs onto its own cuda stream. This can be a user provided stream, similar to how we did it in #64823

Use case

No response

[jinbum-kim]

Hi, I’m interested in NIXL(Dynamo) and have been reading through the related issues and PRs to better understand the context. Is this issue open to community contributions? If so, I’d be interested in working on it.