Near-Real-Time indexes lost when server is stopped

The server does not respect the stop order of the caches dictated by cache manager and tries to stop all caches in random order.

When a cache is indexed and uses the "near-real-time" index manager, it relies on the indexing caches being available during shutdown of the data cache so that it can properly flush the index to persistence storage, and this fails due to one or more index caches being unavailable.