Horizon syn correct behavior

[SWvheerden]

Currently on Horizon sync, if the node fails horizon sync, the node can be stuck in a broken state if it cannot find a peer with the exact same chain. This can leave the peer in a broken state still a user manually deletes the base node's database.

This issue is to discuss ideas

[SWvheerden]

The correct goal should be that when a peer exit's horizon state, the node is in the best state possible.

So when a node finds some issue with the kernel sync, it should rewind the kernels back to the starting height.
Then exit horizon sync and let listing mode + header sync decide the correct approach with a new(or same peer if the peer was not banned) peer.

[SWvheerden]

Currently utxo sync is not resumable, which it should be. The utxo set is downloaded in go , but we should be able to make it so that it can resume.

[hansieodendaal]

Fixing system-level use for #6006