reflect image-pull issues into member status

[joel-bluedata]

This is a sequel to the work done for issue #562.

A pod might be blocked from coming up if the image pull is taking a long time, or the image is bad (nonexistent), or the image repo is bad (nonexistent or unreachable). It would be nice if we could help kdcluster users see this without requiring them to dig around in pod status or events.

Unfortunately we probably can't do much more than tell them "right now this member is busy trying to get its container image pulled" and perhaps also "here's how long it's been doing that".