osquery instance should not restart when the KATC configuration changes

[RebeccaMahany]

Currently, the osquery runner restarts all instances when Ping is called. (See: https://github.com/kolide/launcher/blob/main/pkg/osquery/runtime/runner.go#L286-L302.) On launcher first-time startup, launcher will receive an updated KATC configuration the first time it talks to the control server. Consequently, on launcher first-time startup, the osquery runner will start an instance, then restart it almost immediately afterward. This immediate restart is problematic because it delays the time before osquery is fully functional and running queries, and additionally because osquery startup is a little fragile on Windows and having to do it not once but twice is not desirable.

This issue is for researching options that will allow us to avoid this restart.

Research notes:

• It doesn't look like we can de-register individual plugins, so I don't think we can de-register the KATC table plugins and then re-register new ones.
• Because the kolide_grpc extension is required (we indicate this with the --extensions_require flag when starting up osquery), I don't think we can safely de-register the entire extension and start up a new one in its place -- but it's probably worthwhile to validate this assumption.
• We cannot register plugins after the server has started, so we cannot add the new KATC tables when we receive them after osquery instance startup.
• Therefore -- we have to either wait for the control server data to become available, or ensure that the KATC configuration is available ahead of time or on demand.
• We could make the config available ahead of time by packaging launcher with the current KATC config. (This wouldn't help with organizations that continue to use older installers, though.)
• We could make the config available on demand via e.g. unauthed K2 API endpoint.