Handle http endpoint ListUnderReplicatedLedgerService in twitter's way.

while in codereview of http endpoint -- #520 , there was good suggestion from @yzang to make ListUnderReplicatedLedgerService better.

> If we have a large number of underreplicated ledgers, this will take a long time to run. What we do in Twitter is to run the underreplicated ledger analyzer periodically in auditor and store a list of underreplicated ledgers, so Http Endpoint will just return the list directly.

Start with the ListUnderReplicatedLedgerService HTTP endpoint and the related code review in #520. Review how the auditor periodically runs the underreplicated ledger analyzer and stores its list; done means the endpoint returns that stored list directly instead of rerunning a potentially long analysis.

Written by the indexing model from the issue text.