chore: refactor Datadog Status/Flare component in preparation for shared telemetry via reflector

[tobz]

Summary

This PR does a bunch of refactoring around how we create/spawn the component previously known as the "Datadog Status/Flare Destination".

Primarily, it moves the code into the ADP binary crate itself, and de-componentizes it: now we just have standalone types, we spawn a standalone task, and so on. This is in preparation of having it draw its telemetry data from a new "reflector"-based pattern. We've also done some other work around ListenAddress to make it possible to cleanly pull the "what is our secure API listen address?" information necessary when registering ourselves as a remote agent.

This PR also includes the basic code for the reflector, which is included to intend to show roughly what we'll be moving to. This PR doesn't include all of the work to actually utilize the reflector, though, which will be part of a follow-up PR.

Change Type

• Bug fix
• New feature
• Non-functional (chore, refactoring, docs)
• Performance

How did you test this PR?

Built and ran ADP locally in non-standalone mode, and ensured that the Remote Agent task that replaces the component still runs and registers ADP as a remote agent.

References

N/A

[rayz]

LGTM, just curious: are there any existing code we could migrate to use this new reflector pattern besides the shared telemetry?

[tobz]

LGTM, just curious: are there any existing code we could migrate to use this new reflector pattern besides the shared telemetry?

My thought is:

• status/flare support
• streaming telemetry API (gRPC stuff for a theoretical admin UI showing live stats)

We could potentially also use it for internal telemetry, like the Prometheus scrape endpoint... but it's not clear to me (yet) if all of those use cases can be aggregated the same way, since the implication is that you need the reflector to do aggregation for metrics so you always get the latest gauge value, or the counter total, and so on.

The other thing is that we can use the reflector pattern for logs, too... where it might just be that we have a fixed-size ring buffer of the logs instead of an aggregated set of metrics, etc.