Response parsing for projects ending in "v1", "beta" or "alpha" causes image lookup to fail

[pwschuurman]

We have a function to "clean" a link fetched from the GCE API. This is typically a fully qualified link like the following:

https://www.googleapis.com/compute/v1/projects/k8s-jkns-gci-gce-alpha/global/images/gcepd-csi-e2e-f608826a-5761-4730-ba31-fabbc5c2ef08

The regex is not strict enough to segment based on "/" boundaries, so if the project ends in alpha (in this example), it will lead to the "clean" functionality to truncate the "projects/" segment, resulting in a parse error later.

See this gist for an example e2e test failure output: https://gist.github.com/pwschuurman/b5030e08538828514fe9c045a28df3e8. The following is a link to a prow test failure: https://prow.k8s.io/view/gs/kubernetes-jenkins/pr-logs/pull/kubernetes-sigs_gcp-compute-persistent-disk-csi-driver/1542/pull-gcp-compute-persistent-disk-csi-driver-e2e/1735154233054859264

[pwschuurman]

We've seen this type of bug before in #1302