[maps] ignore indices without geometry field in vector tile requests

[nreese]

Solutions use wide data views that access many elasticsearch indices, for example metrics*. This can cause shard failures for indices that do not contain geospatial field. These shard failures are a nuisance to users and can be removed by adding an exists filter to the query. This is similar to #73304. #73304 has been resolved for search source data requests. The issue is not resolved for vector tile data requests.

Steps to view problem

1. In console, run:

   PUT geo1
   {}
   
   PUT geo1/_mapping
   {
     "properties": {
       "location": {
         "type": "geo_point"
       }
     }
   }
   
   PUT geo1/_doc/1
   {
     "location": "25,25"
   }
   
   PUT geo2
   {}
   
   PUT geo2/_doc/1
   {}
   

2. Create geo* data view
3. create new map
4. add documents layer from geo* data view.
5. Add heatmap layer from geo* data view.
6. notice how layer indicates warning because geo2 index could not return results.
   
7. Switch scaling to limit to 10000. Notice how warning no longer exists. This is because bounds filter for geojson request includes clause to ensure geometry field exists. Vector tile requests should do the same thing.

[elasticmachine]

Pinging @elastic/kibana-presentation (Team:Presentation)