Fix read_schema function for multi-column foreign key constaints

[andrew-farries]

Update the query used by read_schema to fetch foreign key information. This fixes the problem of columns and referencedColumns containing duplicate entries for multi-column foreign keys.

For a foreign key defined like "fk_sellers" FOREIGN KEY (sellers_name, sellers_zip) REFERENCES sellers(name, zip) the change in behaviour is as follows:

Before:

{
      ...
      "foreignKeys": {
        "fk_sellers": {
          "name": "fk_sellers",
          "columns": [
            "sellers_name",
            "sellers_zip",
            "sellers_name",
            "sellers_zip"
          ],
          "referencedTable": "sellers",
          "referencedColumns": [
            "name",
            "zip",
            "name",
            "zip"
          ],
          "onDelete": "NO ACTION"
        }
      },
}

After:

{
      ...
      "foreignKeys": {
        "fk_sellers": {
          "name": "fk_sellers",
          "columns": [
            "sellers_name",
            "sellers_zip"
          ],
          "referencedTable": "sellers",
          "referencedColumns": [
            "name",
            "zip"
          ],
          "onDelete": "NO ACTION"
        }
      },
}

The solution here is to perform grouping and array aggregation of the referencing table's columns before the join to the referenced table's columns.

Fixes #475

[exekias]

nice!