specifically not doing this for aaa since I haven't touched it all

This uses an ugly workaround but the theory is it's hidden in this
module so may it's OK for consumers.

I suspect I'm going to have to tune prow-build's nodepool to use
UBUNTU_CONTAINERD instead of COS for the benefit of kind ipv6 tests

Create a k8s-infra-gke-project module to manage the project that hosts
the cluster. Since I deployed the prototype prow-build-test cluster in
the kubernetes-public project, I based this module off of what looked
relevant from `ensure-e2e-project.sh`

the daemonsets are to tune the build cluster for the benefit of tests
that rely on kind

I tried switching boskos-metrics to gce ingress but it expects any backed
service to 200 OK on /, while the prometheus metrics endpoint exposed as
boskos-metrics 404's on anything but /metrics

kubetest defaults to asking for a project of type `gce-project` if it's
not explicitly specified via --gcp-project-type=. So initially I
thought it would be a good idea to use that for the bsokos pool in this
build cluster.

The problem is when we try to use monitoring.prow.k8s.io's boskos
resource dashboard to show this boskos' resources.  There's nothing to
differentiate from k8s-prow-builds' gce-project pool, and this boskos'
gce-project pool.

The easiest thing to do for now is to use uniquely named pools across
boskos instances. This will require that jobs explicitly add a flag
--gcp-project-type=k8s-infra-gce-project when migrating to this
build cluster

use a smaller machine type and standard disk, to match what is
currently used by k8s-prow/prow, which is the prow.k8s.io
trusted cluster (but also the service cluster)

for extra paranoia I have named the cluster and associated
service accounts differently, so that somebody pointed at the wrong
project couldn't accidentally the wrong cluster

service accounts:
- trusted cluster jobs have yet to require boskos so we're not going to
  make a service account for its janitor
- allow jobs in this cluster to use the k8s-infra-gcp-auditor service
  account

created 2 manual projects and 40 boskos projects

actuated via:
- terraform destroy before removing the .tf files
- gcloud projects delete for the e2e projects
- gcloud compute addresses delete for the boskos-metrics ip

make a projects dir and move all projects into it, instead of having
modules be "the one dir that isn't a project"

update README to follow suit

update READMEs accordingly

this has been broken out into a separate issue