The elastic doc manager should be able to retry on connection failures

[travisp]

If there is a temporary connection failure to elasticsearch, an insert, update, or delete will be lost permanently. Sometimes the only way to get back to correct data consistency is to restart with a new index, which is fairly inconvenient. This was previously discussed in the mongo connector -- yougov/mongo-connector#479 -- but the issue was closed, with the conclusion that each individual doc manager should handle the retrying logic.

Example fix in the postgresql doc manager:

cureatr/postgresql-doc-manager@4d333d6