Diagnose Kaniko pod failures.

A user reported via [slack](https://kubernetes.slack.com/archives/CABQMSZA6/p1603309658143900) that they were not able to get more information on why the kaniko pod failed.
Skaffold tries to fetch the pod logs, however, the pod could fail even before logs. Hook with `diag` pkg to monitor the kaniko pod and report error events to use when no logs are available.

If `skaffold` binary is run on a container, do not show the survey prompt.
```
Building [rg.nl-ams.scw.cloud/njkonl/covid19-per-capita]...
time="2020-10-21T19:42:57Z" level=debug msg="no kube-context set and no kubeConfig found, attempting in-cluster config"
time="2020-10-21T19:42:57Z" level=info msg="Waiting for kaniko-75g8p to be initialized"
time="2020-10-21T19:42:59Z" level=debug msg="Running command: [kubectl --context exec -i kaniko-75g8p -c kaniko-init-container -n g8s -- tar -xf - -C /kaniko/buildcontext]"
time="2020-10-21T19:42:59Z" level=debug msg="Found dependencies for dockerfile: [{requirements.txt /usr/src/app true} {. /usr/src/app true}]"
Help improve Skaffold with our 2-minute anonymous survey: run 'skaffold survey'
```