feat: NTX Actor support for notes that predate account creation

[sergerad]

Context

There is an edge case whereby transactions that relate to not yet existent network accounts are never processed by the NTB. See #1317.

Closes #1317.

No changelog as this is being merged into tracking branch.

Changes

• Move mempool events to Arc to reduce cost of cloning.
• Update NTB Coordinator to determine which actors should receive which events rather than broadcasting to all.
• Update NTB Coordinator to keep track of mempool events with transactions that predate corresponding network accounts.

[sergerad]

@bobbinth this is probably not sufficient. I suppose I need to search the prefixes involved within every note of the transaction? Not just the one pertaining to the account delta?

[bobbinth]

Yes, in fact, I don't think the account from account delta is relevant in this context. Basically, for every transaction we need to get a list of output notes (we only care about network notes addressed to a single account), and then route them to their respective accounts. The notes that remain "unassigned" after this process should be treated separately (i.e., put into some pool from which they can be assigned to an account if/when it gets created).

[sergerad]

The model we have right now is that we broadcast every event to the actors. Initially we did instead filter out irrelevant actors but we removed that logic for some reason (lost in sea of resolved comments in first PR).

This commit shows the logic we removed. Essentially:

    fn find_affected_accounts(
        account_delta: Option<&AccountUpdateDetails>,
        network_notes: &[NetworkNote],
    ) -> HashSet<NetworkAccountPrefix> {

We might want to go back to this both for this PR and in the future if we decide to only spawn ephemeral tasks rather than keeping a task around for every actor.

[Mirko-von-Leipzig]

Not a full review, but one issue I did pick up which needs thinking about

[Mirko-von-Leipzig]

I think this works.

I would like us to work on simplifying the overall flow somehow. I've been thinking about what assumptions we can make to do this and I have an idea I'll write up soon.