Make the API queries more robust in case of a network restart

[Leo-Besancon]

Some API queries do not respond the expected result in case we are waiting for a network restart.
For example, the get_stakers() list is empty if we query a cycle that is no longer in the pos_history. We should query the stakers list with the cycle of the last_start_period in that case, not the cycle associated with the current time.

[aoudiamoncef]

Hi @Leo-Besancon

Where could we found last_start_period ? in gRPC side, we have a dedicated filed called context which gives information about the execution of the query, we could add a message or anything to inform the caller that's It's not the current cycle but a previous one etc

[Leo-Besancon]

Hey @aoudiamoncef! I'm not 100% sure yet, but I think this should not be resolved at the API level directly. It's more that the info queried by the API should be more resilient.
For example, the get_stakers api point calls get_cycle_active_rolls, that calls get_all_roll_counts(cycle) in the pos_final_state.rs file.
I think it's this last function (that has access to the last_start_period from the final_state) that should handle this case :)