Fixes the test generation for the etl task

[jasminenoack]

Addresses #468

This changed the parsing for etl to handle the new cannonical structure which looks like:

{
      "uuid": "78a7a9f9-4490-4a47-8ee9-5a38bb47d28f",
      "description": "single letter",
      "property": "transform",
      "input": {
        "legacy": {
          "1": ["A"]
        }
      },
      "expected": {
        "a": 1
      }
}

It removes the nested legacy key in the input section so that the json object for the case will match the expectations in the parser. The edit_etl_case returns the following pattern:

{
      "uuid": "78a7a9f9-4490-4a47-8ee9-5a38bb47d28f",
      "description": "single letter",
      "property": "transform",
      "input": {
        "1": ["A"]
      },
      "expected": {
        "a": 1
      }
}

Once the generator was working, I found an additional bug in the template that it was looping over the cases multiple times so I remove the inner loop so the 4 cases would only appear once. This appears to have been a copy/paste mistake from generators that do use nesting.