Support persistent connections to App Search

[jsvd]

Currently the Client class creates a new HttpClient per request, which means a new connection as well.

For high load scenarios, the object allocation and round trip cost is too high, so the Client class should contain a single httpclient and manage its lifecycle.

This would also create the requirement to ensure the class behaves in a threadsafe way, so we can create a single client and share it across multiple indexing threads.