Spatial filters stomp on existing filters in ElasticSearch backend

[joshdrake]

Narrow filters are applied before the spatial queries using a query filter. If spatial filtering is performed, these filters are stomped on. This can be easily reproduced with a basic case:

ninth_and_mass = Point(-95.23592948913574, 38.96753407043678)
max_dist = D(mi=2)
sqs = SearchQuerySet().models(Shop).dwithin('location', ninth_and_mass, max_dist)

This will generate a filter fragment looking like this:

{
    "filtered": {
        "filter": {
            "geo_distance": {
                "distance": 3.218688, 
                "location": {
                    "lat": 38.967534, 
                    "lon": -95.235929
                }
            }
        }
    }
}

... when it should be this:

{
    "filtered": {
        "filter": {
            "and": [
                {
                    "fquery": {
                        "_cache": true, 
                        "query": {
                            "query_string": {
                                "query": "django_ct:(shops.shop)"
                            }
                        }
                    }
                }, 
                {
                    "geo_distance": {
                        "distance": 3.218688, 
                        "location": {
                            "lat": 38.967534, 
                            "lon": -95.235929
                        }
                    }
                }
            ]
        }
    }
}

I've submitted a pull request to correct this issue.

[jezdez]

I've pulled the first commit, please feel free to open a new pull request about d5cc42f.