[Corollary to #4161] Redis Connection Check when Using Sentinel

[dstarner]

Fixes: #3985, #3984, and a bug introduced in #4161

Redis Sentinel support was adding to the settings and configuration, but the extras AppConfig was not updated to account for Sentinel when starting the application. Without this fix it shouldn't cause the application to crash, but to throw a false positive connection issue. This fixes the check to account for when sentinel is configured

[lampwins]

I would actually be okay with ripping this check out. It isn't necessarily adding any value over a sanity check on application upstart. For instance, it does nothing to check that the connection is still open over time. Cacheops is already set to safely degrade if the connection fails and error messages will be logged for failed webhook requests on account of the redis connection.

@jeremystretch thoughts? This would also resolve #3985

[dstarner]

I feel like if a check is ever needed (which i believe it shouldnt be) then it also shouldn't be in a blocking path that prevents local-only commands from running, like mentioned in #3985. If it's believed that the check can be fully removed, I don't mind reworking this PR to just remove the original check.

[dstarner]

Or, I guess if we wanted to keep the check, in a similar fashion to IS_TESTING in the configuration, we could add either a REQUIRES_REDIS or LOCAL_ONLY variable that wraps the check...but that seems even hackier and it has a slight code smell to it.

[jeremystretch]

I have no objection to removing the check entirely. I agree that it is not necessary.

[jeremystretch]

We probably do want to add some error handling around webhooks at least, though. Without this check in place, taking any action which results in the queuing of a webhook will cause a server error if the Redis service is unreachable. I'm happy to work on that; just wanted to make a note of it here.