fix: Check all containers and init containers for proxy before recreating pod (#662)

Since the controller is now using sidecar init containers, these should be checked for existence when determining if a pod needs to be recreated because the proxy injection failed. In other words, currently we get false positives if the main workload is erroring and the proxy was injected as a sidecar init container, which causes the controller to remove the pod instead of letting it progress.

I ran into some errors running the go_lint step in the Makefile (even before making changes) and ran into a problem during the terraform execution of building the infra for the e2e tests, so I wasn't able to complete them as laid out in the docs. I did however remove the go_lint step, build the controller image and replace it in my cluster and confirmed the desired effect.

Co-authored-by: Jonathan Hess (he/him) <103529393+hessjcg@users.noreply.github.com>

Author: zstevenson-payments

internal/workload/podspec_updates.go

@@ -226,7 +226,9 @@ func (u *Updater) CheckWorkloadContainers(wl *PodWorkload, matches []*cloudsqlap
 	for _, p := range matches {
 		wantName := ContainerName(p)
 		var found bool
-		for _, c := range wl.PodSpec().Containers {
+		podSpec := wl.PodSpec()
+		allContainers := append(podSpec.Containers, podSpec.InitContainers...)
+		for _, c := range allContainers {
 			if c.Name == wantName {
 				found = true
 			}