chore: simplify actor start up

[SantiagoPittella]

closes #1888

Simplifies NTX actor startup by removing the AccountOrigin enum. Actors now take a plain NetworkAccountId and on startup wait for a committed account state in the database before producing transactions.

The coordinator can now blindly spawn actors without caring about where the account state comes from since the actor handles it internally.

Regarding monitoring

cc @sergerad @Mirko-von-Leipzig

Additionally, and did not commit it, to monitor the actgor's spawning I added a local observability setup using Grafana and Tempo, using the already exported traces from the node.

This adds a docker-compose.yml that spins up Tempo as the trace backend and Grafana as the dashboard frontend. Tempo receives traces on port 4317 and Grafana is pre-configured with Tempo as a datasource so there is zero manual setup.

This is useful for debugging NTX actor behavior during development since you can see the full span tree for actor spawns, transaction execution, block commits, and gRPC calls; and also Tempo can derive rate, error rate, and duration metrics from traces, which was really useful to track actors spawning.

I can create a PR with that setup if you guys consider it worth it.

[Mirko-von-Leipzig]

@SantiagoPittella I think it makes sense to add those - I imagine to the node's compose file that @sergerad is adding.

There should be an open issue about adding more stuff to the compose file.