feature/batches execution

[baywet]

• adds an execute method for JSON batches to avoid customers juggling with clients
• reverts string serialization bypass to avoid unconsistent behavior
• adds defensive programming - switches to json objects to avoid too much string manipulations - fixes a bug where the service root would be appended twice - adds missing unit tests

This PR not only improves code coverage and reliability of the whole batching implementation, but also makes usage simpler.
https://github.com/microsoftgraph/microsoft-graph-docs/compare/feature/java-sdk-v3?expand=1

[tbelaire]

Just as a note, if you're redoing how this batch request works, I would find it much more useful if it didn't reference okhttp Requests, as I've already hooked up our own IConnection and IHttpProvider, and don't really want to use okhttp.

What would be ideal from my perspective would be a BatchRequest extends BaseRequest with a method like public JsonObject post(BatchRequest batchRequest), where the BatchRequest is some trivially serializable to Json with a list of BatchRequestItems. But these BatchRequestItems could be subclassed for each request type when generating the rest of the classes automatically:

  class UpdateMessageBatchRequestItem extends BatchRequestItem {
    @Expose public Message body;
  }

Then IMessageRequest could expose toBatchRequest which would mean that you could build up these
batch requests in the same way as you'd use the rest of the GraphApi sdk.

Sorry if this feedback comes too late, or isn't too useful here.

[baywet]

@tbelaire I think this is very valuable feedback that is unlocked by the fact that all the base requests used to be in the service library and now are in this core library.
At this point I wasn't looking at rebuilding the whole batch constructs, I only fixed issues that were obvious and added a superset method to facilitate usage and avoid juggling with clients.

I don't want to block the v3 release on that and I want to involve the program managers in that potential redesign. I agree with the fact that the current implementation is overly complex and reimplements things we should get from the httpprovider and other layers. Do you mind opening a new issue providing all the details you already provided please?