Loader should not call the "list" Iglu endpoint for Snowflake/Databri…

…cks tables

When loading to Redshift, the loader needs to call the "list" Iglu
api endpoint in order to discover all minor versions of a schema. It
uses the list to merge the schemas.

But for Snowflake/Databricks tables the loader does not need to merge
schemas, so there is no need to call the "list" endpoint.

In RDB Loader version 6.0.0 we accidentally changed the loader so it
calls the "list" endpoint regardless of the table type. This is a
problem for Snowflake/Databricks loads because:

- it is inefficient to call a api endpoint when we don't need the result
- some old Iglu repos do not support the list endpoint.

modules/loader/src/main/scala/com/snowplowanalytics/snowplow/rdbloader/discovery/DataDiscovery.scala

@@ -181,7 +181,10 @@ object DataDiscovery {
                                .leftMap(er => LoaderError.DiscoveryError(IgluError(s"Error inferring columns names $er")))
                          }
                      }
-      models <- getShredModels[F](nonAtomicTypes)
+      models <- message.typesInfo match {
+                  case TypesInfo.Shredded(_)   => getShredModels[F](nonAtomicTypes)
+                  case TypesInfo.WideRow(_, _) => EitherT.rightT[F, LoaderError](Map.empty[SchemaKey, DiscoveredShredModels])
+                }
     } yield DataDiscovery(message.base, types.distinct, message.compression, message.typesInfo, wideColumns, models)
   }