Fix errcheck in service/worker/

[MichaelSnowden]

What changed?
We now surface errors from code within service/worker.

1. We log an error when we fail to unregister a membership listener in the per-namespace worker
2. We log a fatal error when we fail to register the membership listener during startup
3. We log a fatal error whenever any SDK workers fail to start
4. The scheduler will retry now whenever the side effect of creating a new UUID fails (this should be rare, but we still shouldn't ignore these errors).

Why?
I made these changes so that we don't silently swallow these errors. Additionally, this will enable us to run errcheck on this directory.

How did you test it?
I tested this using the CI. However, I made sure that the not-so-important errors, like unregistering a membership listener, are only logged, not treated as fatal.

Potential risks
This could cause errors on startup/shutdown that users aren't used to.

Is hotfix candidate?
Let's note it because OSS users might see a new error and wonder why.

[dnr]

I appreciate the effort but this is wrong: returning an error here will end the workflow. Errors from terminate or cancel are supposed to be swallowed since they'll be retried at the next action attempt.

And most of the other changes in this file are just unnecessary clutter for type errors that can't ever happen.

Can we revert at least the changes to this file?