Improve logging on autodiscover recoverable errors

[jsoriano]

There are some cases where autodiscover configuration fails, but the failure is related to some internal state that is automatically recovered after some retries.

This happens for example with #11834. There filebeat tries to reopen a file that hasn't been completely closed after being used in a previous configuration, and logs errors like this one:

Error creating runner from config: Can only start an input when all related states are finished.

Filebeat automatically recovers from these errors, so they shouldn't be logged at the error level, as this causes confusion when investigating other issues.

[elasticmachine]

Pinging @elastic/integrations-platforms (Team:Platforms)