Should retry on API errors like 5xx

[butenkor]

Affects all modules (should be covered by JVM SDK)

[ghost]

Proposed solution:

• provide utility function which wraps a sphere client in RetrySphereClientDecorator
• use that function in a services' constructors (cause they are responsible for performing CRUD operations on CTP and they take a sphereClient as a parameter)

@butenkor I'm not sure about the strategy regarding calculation of waiting time between subsequent retry attemps. Could you provide me expected numbers so that it was reasonable?
E.g. Would a sequence 2s, 2s, 2s, 4s, 4s, 4s, ... , 512s, 512s, 512s of waiting time between following attempts be ok? It would result in 27 attempts, lasting almost 2 hours in a case of the constant failure.

[heshamMassoud]

Doesn't have to be a responsibility of the sync library. Showing an example in integration tests should be enough.