Refactor round tracker to not block service startup

[bajtos]

When Glif API performance is degraded, we are not able to deploy new versions of spark-api because the startup times out while fetching RoundStart logs from the chain history.

Let's redesign our code to initialise the round tracker in background and no longer block service startup.

[juliangruber]

Adding more context:

• @bajtos: "We can also refactor the startup code to try to check if we are already tracking the current round in the database, I had that in mind, but never found time to implement it."
• I recently increased the block count for the events query from 250 to 500, since spark-evaluate had an issue, and the last event was too old. This blocked spark-api from starting up, as it couldn't find any event. The query shouldn't be too big, as it causes performance degradation from Glif. Idea: add an exponential increase there (check 10 blocks, then 100, etc)

[bajtos]

We have already made quite a few improvements in this area. See the list of PRs linked to this issue.

The next step is to assess the current situation, identify the remaining conditions in which the service startup is blocked, and propose an implementation plan.