GPKG working copy adds auto_int_pk for no reason if pk column isn't first column

[olsen232]

GPKG requires that tables have a single integer primary key.
GPKG doesn't require that that PK is in the first column.

If a dataset has data that doesn't conform to the GPKG requirement, we demote any of the PK columns that it does have to be not PK columns, but only UNIQUE NOT NULL, and we add a new PK column called auto_int_pk that conforms to GPKG spec.

We shouldn't be doing this in the case where the dataset already conforms to the GPKG requirement, even if the PK column is not the first column. The issue is that we weren't recognizing datasets as conformant when the PK wasn't the first column, and so were adding in the auto_int_pk column. More confusingly, the code that hide edits that were made automatically and necessarily by Kart, wasn't hiding this added column, so you would see spurious diffs like the following on a fresh checkout:

--- nz_primary_road_parcels:meta:schema.json
+++ nz_primary_road_parcels:meta:schema.json
  [
+   {
+     "id": "620a30b4-f793-53ff-0f6b-49003d59aa66",
+     "name": "auto_int_pk",
+     "dataType": "integer",
+     "primaryKeyIndex": 0,
+     "size": 32
+   },
    {
      "id": "b918dd54-73dc-4752-894a-1078e7360712",
      "name": "shape",
      "dataType": "geometry",
      "geometryType": "GEOMETRY",
      "geometryCRS": "EPSG:4167"
    },
-   {
-     "id": "e2c7c7c9-c017-4d66-894b-de4c599231e8",
-     "name": "id",
-     "dataType": "integer",
-     "primaryKeyIndex": 0,
-     "size": 32
-   },
+   {
+     "id": "be3f60fb-0890-f9ec-649d-1c41f4da0369",
+     "name": "id",
+     "dataType": "integer",
+     "size": 32
+   },
...