Fix Ingress awaiter for ExternalName Services

[lblackstone]

The await logic was incorrectly waiting for active Pods for all Service
types. However, Services with type: ExternalName do not have
associated Pods and should succeed as long as the Service exists.

Fixes #317

[hausdorff]

A better name for this is now probably something like serviceReady, as Services with type ExternalName don't have endpoints.

[hausdorff]

Arg, dammit. I'm testing out the GH extension and it apparently doesn't let you batch up review comments?

[hausdorff]

It seems like we have to do this also for case <-iia.config.ctx.Done():?

[hausdorff]

if event.Type == watch.Deleted {, don't we want to delete this from iia.externalServices? And now that I think about it, is there a reason we don't do this for iia.endpointExists as well?

[hausdorff]

I think I'm missing something -- seems like it would be better to just get rid of iia.endpointsReady and simply call iia.checkIfEndpointsReady() wherever you would check iia.endpointsReady? Seems redundant to have the function and the variable.

[hausdorff]

Because @lblackstone is not currently here, I've gone ahead and "addressed" my comments. To my eye, this is semantically equivalent, but with the (small) changes noted in the following list. I'd like to turn this sort of thing around pretty quick, so I'll file a follow-up issue to have @lblackstone do a post-commit review to make sure I didn't do anything silly.

• Renamed iia.endpointExists -> iia.knownEndpointObjects and iia.externalServices -> iia.knownExternalNameServices, and made both of them sets.String.
  • Rationale: Fundamentally the awaiter now checks readiness by determining if a v1/Endpoint object exists, but only if the v1/Service is not of type ExternalName. If we maintain a set of Endpoints and Service objects that have type ExternalName, it seems to make it quite simple to check whether this condition is true. We rename because we want the names to reflect this usage.
• Removed iia.endpointsReady, replaced all instances where we check this with iia.checkIfEndpointsReady().
  • Rationale: Unlike iia.ingressReady, this field is now quite complex, and we find ourselves having to update it before every check to ensure correctness. iia.ingressReady, on the other hand, is set only in one place, and is fairly simple. So, we think it makes sense to just call iia.checkIfEndpointsReady() directly.
• When we receive a v1/Event indicating a deletion of a Service or Endpoint, we now delete them from iia.knownExternalNameServices and iia.knownEndpointObjects, respectively. Before, we'd keep them in a map, and set their values to false.
• We un-delete on of the test cases, and simply add the new test case alongside it. (I believe the deletion was originally a mistake.)