Add an option to write the aggregation type in the response

[obourgain]

This pull requests add two optional parameters to the aggregations toXContent, one to write the TYPE of the aggregation and the other to write the classname of the aggregation.

The type field exists for facets but not for aggregations. I believe these informations can be usefull to create mappers without having to either rely on the request or infering the type from the response.

Existing client will not be impacted because the new informations are optional, but they may start to use it.

Suggestions for the fields and parameters names are welcome.
Maybe writing class.getSimpleName() is more robust than class.getName() in regard of refactorings.

It would be wonderful if this could also be merged with branch 1.x.

ouput would be like :

{  
   "_type":"terms",
   "_concrete_type":"org.elasticsearch.search.aggregations.bucket.terms.StringTerms",
   "doc_count_error_upper_bound":0,
   "buckets":[  
      {  
         "key":"val0",
         "doc_count":1
      }
   ]
}

[jpountz]

@obourgain We discussed this issue in #5867 which lead to #6465, which would add the ability to add arbitrary metadata to aggregations and have it back in the responses. It looks like the PR is a bit stalled however.