These variables and outputs aren't used anywhere and can be removed.

The existing version would sometimes panic when creating NLBs and was
also missing a few options specific to NLBs. There are probably many
other fixes and improvements that come with the newer provider.

This gives the installer the ability to tear down NLBs.

We've noticed an elevated rate of installation failures recently. The
root cause appears to be 50-90 seconds of latency added to traffic going
through the internal ELB on port 49500. This was causing Ignition's
connections to timeout, resulting in the machines never provisioning.

AWS's NLBs don't seem to have this high latency, so we've decided to
move over to them instead. With the move to NLBs, we also get the
ability to add individual health checks for each port instead of just a
single health check for each load balancer. Also, NLBs are cheaper.

This commit drops support for ingress and the console. Since the console
and router aren't currently configured correctly, nobody should notice
that this is gone. It was easier to drop support in this commit rather
than continue to try to plumb through the existing implementation
knowing that it was going to have to change in the future. Once the
router has a strategy for ingress, we'll re-add this functionality using
the new NLBs.

This also drop support for the `<cluster-name>-k8s` DNS entry. We aren't
aware of any consumers and it was going to be tedious to keep this
working.