Add GeoInterface to Python MapScript Classes

[geographika]

This pull request adds a __geo_interface__ property for PointObj, LineObj, and ShapeObj.
See https://gist.github.com/sgillies/2217756 for details about the geo_interface.

It allows a GeoJSON-like interface for MapScript objects to help with interoperability with other geospatial Python libraries.

        res = layer.getResult(0)
        s = layer.getShape(res)
        s.itemdefinitions = l.getItemDefinitions()
        print(s.__geo_interface__)

Results in:

          {
            'geometry': {
                'type': 'Polygon',
                'coordinates': [[(-0.25, 51.227222), (-0.25, 51.727222), (0.25, 51.727222),
                                 (0.25, 51.227222), (-0.25, 51.227222)]]
            },
            'type': 'Feature',
            'bbox': (-0.25, 51.227222, 0.25, 51.727222),
            'properties': {'FNAME': 'A Polygon', 'FID': 1}
         }

A new MapScript method getItemDefinitions has been added to the layerObj.

Returns the requested item's field type.
A layer must be open to retrieve the item definition. 

Pass in the attribute index to retrieve the type. The 
layer's numitems property contains the number of items 
available, and the first item is index zero.

This allows the type of an item (field) to be returned as a string e.g. Character, Integer, Real (these are all GML data types as listed at https://mapserver.org/ogc/wfs_server.html).

Types are set on fields in a layer through gml_[item name]_type METADATA settings, or using "gml_types" "auto"

These types can then be set as a property on a shapeObj as shown in the example above so that item values are returned in the GeoJSON output as ints, floats, or strings.

A review of the C code in mapscript/swiginc/layer.i would be very much appreciated.

[sdlime]

I'll review today...

[geographika]

@sdlime - it is mainly the C function at https://github.com/mapserver/mapserver/pull/5765/files#diff-deee1b0c6f077b32fa05657f1799802c that it would be good to review. The Python extension code and tests should be fine. I changed the approach from returning a gmlItemDefinition which involved adding SWIG guards everywhere in the mapows.h and simply return the data type name as a copied string.

[sdlime]

@sdlime - it is mainly the C function at https://github.com/mapserver/mapserver/pull/5765/files#diff-deee1b0c6f077b32fa05657f1799802c that it would be good to review. The Python extension code and tests should be fine. I changed the approach from returning a gmlItemDefinition which involved adding SWIG guards everywhere in the mapows.h and simply return the data type name as a copied string.

@geographika Sorry, sorry... Should check item_list is not NULL before accessing it since msGMLItems() can return if it has issues allocating memory. I think it's fine otherwise.

[geographika]

@sdlime - thanks for the review. I've added in the check. I presume if item_list is NULL then there is no need to call msGMLFreeItems(item_list); and that line can stay in the If block?
If you're happy I can merge so it is included in the 7.4 release.

[sdlime]

@sdlime - thanks for the review. I've added in the check. I presume if item_list is NULL then there is no need to call msGMLFreeItems(item_list); and that line can stay in the If block?
If you're happy I can merge so it is included in the 7.4 release.

Looks good to me - merge on!