Return status and actual timestamp from all environments

[motus]

This is the first step to making trials asynchronous. In this PR we just make sure every Environment method that returns status also returns the timestamp of that status. In the next PR we will get these timestamps and status values from the remote environment.

[bpkroth]

Also in this PR, we store the statuses and timestamps in a new table that will later serve as a scheduling queue.

Not sure I agree with this, but maybe I just am not following where you're heading with it.

I think we need to store two separate things:

1. in progress status events of the trial (e.g., how different environments status change at different times). I think that's valuable for debugging and other stateful reasons (e.g., which environments do we still need to call teardown or cleanup on).
2. the overall status of a trial (e.g., pending, in progress, aborting, succeeded, failed, etc.). This could be used to implement a scheduling queue.

However, if implement this with a single table, then the latter requires messy aggregates over the former to find out the overall status of each current trial.

I think that should probably just be stored as a single column in the trial table and updated explicitly with logic inside mlos_bench.

Thoughts?

[motus]

Thoughts?

I think we are on the same page. This is exactly how I want to implement this. I will remove all DB-related code from this PR and move it to the next one that actually implements the queue - i.e., with the corresponding read method, unit tests, etc.

[bpkroth]

LGTM