[leo_manager] Calling leo_manager_api:get_members_of_all_versions get timeout on leo_gateway

[mocchira]

https://github.com/leo-project/leo_gateway/blob/develop/src/leo_gateway_app.erl#L412

leo_manager_api:get_members_of_all_versions return {error, timeout} for a while.
After several timeouts happened,
leo_manager_api:get_members_of_all_versions did work.

Due to this problem,
Getting leo_gateway's status running take too much time more than previous versions.

[mocchira]

https://github.com/leo-project/leo_redundant_manager/blob/develop/src/leo_redundant_manager_worker.erl#L286
in the above function

timer:apply_after(100, leo_redundant_manager_api, create, [])

are called many times,
this makes leo_redundant_manager very high load and cause {error, timeout}.

We need to reduce this call as little as possible.

[yosukehara]

We'll consider leo_redundant_manager supports mnesia's transaction for updating cluster's members.

[mocchira]

Confirmed.