Retry logic is incorrect

[tintoy]

At present, each time the retry ticker goes off, a signal is queued at the ticker channel and these will back up until processed (this will, over time, lead to a very tight loop of repeatedly making calls the CloudControl API which are likely to still fail with RESOURCE_BUSY, which is not exactly being a good citizen).

Consider replacing the ticker with a timer that is restarted after processing is complete (if retry is requested).

[tintoy]

I suspect this may also be responsible for the intermittently-high CPU usage exhibited when the provider is processing many resources in parallel (realistically, most of the provider's time should be spent waiting for responses from the CloudControl API).