fix: custom query row validation failing when SQL contains upper cased columns

[nj1973]

We found this issue on Oracle where column default to upper case but it could easily be reproduced on other engines by writing SQL with upper cased columns, the columns in the actual SQL engine were irrelevant, it was just the SQL we pass in.

The problem was only seen for custom query validation using --comparison-fields. When using --hash the calculated fields resulted in lower case end column names.

I added tests to protect from regressions for Oracle, SQL Server, Teradata vs BigQuery. The fix was global so should wortk on all engines but I was reluctant to add too many tests seeing as integration tests are already pretty slow.

The fix was very minor, in data_validation/clients.py I convert the Ibis schema to have lower case columns:

-    return client.sql(query)
+    iq = client.sql(query)
+    # Normalise all columns in the query to lower case.
+    # https://github.com/GoogleCloudPlatform/professional-services-data-validator/issues/992
+    iq = iq.relabel(dict(zip(iq.columns, [_.lower() for _ in iq.columns])))
+    return iq

[nj1973]

/gcbrun

[nehanene15]

LGTM!