Use full type name when checking max schema key

[pondzix]

Scenario:

• Input batch contains data using schema, let’s say link_click
• link_click schema is used as a context AND as an entity/self describing event
• We have multiple versions of link_click schema

When translated to the content of shredding_complete JSON message, it would contain Iines like this:

      "types": [
        {
          "schemaKey": "iglu:com.snowplowanalytics.snowplow/link_click/jsonschema/1-0-0",
          "snowplowEntity": "SELF_DESCRIBING_EVENT"
        },
        ....
        {
          "schemaKey": "iglu:com.snowplowanalytics.snowplow/link_click/jsonschema/1-0-1",
          "snowplowEntity": "CONTEXT"
        }
      ]

For such scenario it looks like we skip necessary warehouse migration for self describing column. We only execute migration for the context:

INFO Migration: Migrating contexts_com_snowplowanalytics_snowplow_link_click_1 AddColumn(Fragment("ALTER TABLE atomic.events ADD COLUMN contexts_com_snowplowanalytics_snowplow_link_click_1 ARRAY"),List()) (pre-transaction)

but never for unstruct_event_com_snowplowanalytics_snowplow_link_click_1, which results in an error when inserting data to the table:

ERROR Error executing transaction. Sleeping for 30 seconds for the first time
net.snowflake.client.jdbc.SnowflakeSQLException: SQL compilation error: error line 1 at position 1,779
invalid identifier 'UNSTRUCT_EVENT_COM_SNOWPLOWANALYTICS_SNOWPLOW_LINK_CLICK_1'

It seems to be caused by this line where we group incoming types by the name and find max schema key per group. But name doesn’t contain unstruct/context prefix, so it’s possible to “lose” type when schema is used as both context/unstruct + with older version.

I think the solution here could be using full type name instead of simple short name. Full name takes into account if it’s context/unstruct, still groups types, but with additional prefix.