Seiso should fail the pipeline on errors

[bittner]

Currently, when seiso fails on certain errors the pipeline, e.g. on GitLab, succeeds happily.

Instead, it should fail and make the pipeline error out, as we would expect it to do.

Examples

On Git-related errors (corrupted file system or Git history):

 $ seiso image history ${IMAGE_NAME} --keep 15 --delete
Error: retrieving commit hashes failed: object not found

On Kubernetes integration issues (unknown namespace as GitLab chooses it automatically when not specified):

 $ seiso image history ${IMAGE_NAME} --keep 15 --delete
Error: could not retrieve image stream 'foo-9-integration/myimage': imagestreams.image.openshift.io "myimage" is forbidden: User "system:serviceaccount:some-foo-production:gitlab-ci" cannot get imagestreams.image.openshift.io in the namespace "foo-9-integration": no RBAC policy matched