Handle persistence failure better

[iafuture]

Currently, we just hard-stop the process if we detect persistence failure:

    return _persistence->flush()
        .then([resp=std::move(resp)] (auto&& flushStatus) mutable {
            if (!flushStatus.is2xxOK()) {
                K2LOG_E(log::skvsvr, "Persistence failed with status {}", flushStatus);
                // TODO gracefully fail to aid in faster recovery.
                seastar::engine().exit(1);
            }

            K2LOG_D(log::skvsvr, "persistence flush succeeded. Sending response to client");
            return seastar::make_ready_future<std::tuple<Status, ResponseT>>(std::move(resp));
        });

We should handle this more gracefully, probably going via Monitor to the CPO so that we can take cluster-wide action if needed