Fix timeouts not working when not LEADING or FOLLOWING

[tylerkaraszewski]

Details

When a server is neither LEADING nor FOLLOWING we don't check for command timeouts. Normally this doesn't matter, because a server should become LEADING or FOLLOWING very soon, but we run into this when shutting down clusters in tests. If you are the last node left in a cluster, you won't have quorum, and thus cannot process commands.

This is why we've changed BedrockTester to take a timeout in the same PR, because we do this in auth:
https://github.com/Expensify/Auth/blob/0751f4eefd49945a009cc219152fe3b51a684502/test/test/AuthTester.cpp#L219

Because when this is done via Query command (as it is for HC-Tree) we can send a Query to a node that is never able to process it, and the node will never shut down.

I think in production, this is relatively unlikely to occur, but I still think it's "more correct" to honor timeouts in this state.

There are some particulars here about not having a DB handle when we do this, which is generally required to timeout commands because we attach a CommitCount message to even timed out command responses. This skips attaching that line in this particular case, and changes how some parameters are passed/stored/etc in BedrockCore to accomodate not always having an SQLite object available.

Fixed Issues

Found while working on https://github.com/Expensify/Expensify/issues/337537

Tests

Existing

---

Internal Testing Reminder: when changing bedrock, please compile auth against your new changes

[pecanoro]

Manually commenting since the deploy script is broken, this was deployed to webrock