Fix bug with defining connect healthcheck

[lkysow]

Previously we were setting the alias_service field of alias checks to
the service name instead of the service id. This check would then pass
because Consul would consider an alias for a non-existent service id to
be okay. With hashicorp/consul#7384, now the
check will fail if the service id doesn't exist and so those connect
services will be considered unhealthy and be unroutable.

[lkysow]

I've tested with consul from master that has hashicorp/consul#7384 merged and this check is now passing correctly. Without this patch all connect services will be failing and unroutable.

[lkysow]

lkysow/consul-k8s-dev:jun04-svc-alias can be used with lkysow/consul-jun4 if you want to test

[lkysow]

To reproduce:

1. Install consul with lkysow/consul-jun4 consul image
2. Deploy a connect service
3. Notice that it's failing its alias check
4. If you deploy another service and try to make a call between the two they'll fail.
5. If you dig into the envoy logs at debug level you'll see it's not healthy

Then update consul-k8s image to lkysow/consul-k8s-dev:jun04-svc-alias and delete the pods so they redeploy. The check now passes.

[ishustava]

This looks good! My tests were all successful.