Fix error reporting

[lblackstone]

Proposed changes

• Pod errors were logged during updates, but not reported properly
  at the top level if the update failed
• Wrap update error messages to provide better context for
  failures
• Properly extract suberrors from wrapped errors so that they are
  reported

Here's what a failed Pod create looked like before:

  kubernetes:core:Pod (pod-test):
    error: Plan apply failed: 1 error occurred:
    	* resource pod-test-4e7xarx6 was successfully created, but the Kubernetes API server reported that it failed to fully initialize or become live: Resource operation was cancelled for 'pod-test-4e7xarx6'

With this change:

  kubernetes:core:Pod (pod-test):
    error: Plan apply failed: 2 errors occurred:
    	* resource pod-test-33ffnqif was successfully created, but the Kubernetes API server reported that it failed to fully initialize or become live: Resource operation was cancelled for "pod-test-33ffnqif"
    	* containers with unready status: [nginx] -- [ErrImagePull] manifest for nginx:1.13-fail not found: manifest unknown

Here's what a failed update looked like before:

  kubernetes:core:Pod (pod-test):
    error: Plan apply failed: 2 errors occurred:
    	* Resource operation was cancelled for 'pod-test-y8znfbm5'
    	* containers with unready status: [nginx] -- [ImagePullBackOff] Back-off pulling image "nginx:1.13-fail"

With this change:

  kubernetes:core:Pod (pod-test):
    error: Plan apply failed: 2 errors occurred:
    	* the Kubernetes API server reported that "pod-test-xn9sqrod" failed to fully initialize or become live: Resource operation was cancelled for "pod-test-xn9sqrod"
    	* containers with unready status: [nginx] -- [ImagePullBackOff] Back-off pulling image "nginx:1.13-fail"

Related issues (optional)