Fix the MOVE and CANCEL behaviours on Assets

[ripienaar]

Proposed change

There are several consistency and behaviour issues with moving assets:

• It's quite brittle and bandwidth expensive
• If it fails or get cancelled the stream might end up with the incorrect configuration - tags that does not reflect where the stream actually runs
• The cancel API is poorly implemented and does not respond with useful data
• The cancel API is not in the user account

There are some back story in nats-io/natscli#724 (comment)

Dereks suggest we might pick 1 node in the target tag/cluster and bring it up to current, then make that the leader and only then expand the replicas in the target location so they sync from the leader that is then in their cluster. This would greatly improve the performance and cost of super cluster moves.

This might be a bug fix in 2.11.x

Use case

We want to build tooling that proactively balances clusters using moves and cancels, users need to be able to do so manually without fear.

Contribution

No response