standardize how we return schemas from the api

[big-andy-coates]

Newer code tends to return schemas through the API using SQL, e.g.

ID INT KEY, NAME STRING, ADDRESS STRUCT<HOUSE_NO INT, STREET STRING, ...>

Where as the older code exposes this using a json version of the Connect schema, e.g.

[
   {
      "name": "ID",
      "schema": {
         "type": "INT",
        "fields": null,
        "memberSchema": null
       }
   },
  {
      "name": "NAME",
      "schema": {
         "type": "STRING",
        "fields": null,
        "memberSchema": null
       }
   },
  {
      "name": "ADDRESS",
      "schema": {
         "type": "STRUCT",
        "fields": [
             {
                "name": "HOUSE_NO",
                "schema": {
                "type": "INT",
                "fields": null,
                "memberSchema": null
                }
          } ,
          {
                "name": "STREET",
                "schema": {
                "type": "STRING",
                "fields": null,
                "memberSchema": null
                }
           }
        ],
        "memberSchema": null
       }
   }
 
]

The later being handled by EntityUtil.

We should probably standardize on one format.

Note: #3526