Implement automatic retry on certain blockchain errors

[chuck-h]

Background
Telos blockchain transactions can be rejected if they take too much CPU time to execute, and this is often a transient phenomenon (e.g. due to cache misses in a processing node). An immediate retry is usually successful.
Also, api servers sometimes go offline or become congested. In these cases the best strategy is to retry with a different endpoint.

Suggestion
Implement a uniform mechanism to respond to transient api problems by retrying in an appropriate way.