Fix key error and cross-join error during qualx evaluate

[leewyang]

This PR fixes some bugs with the evaluate command in the internal CLI:

• key error when all stages are unsupported.
• unexpected cross-join in per-app results under specific conditions (appName and scaleFactor are all the same, but description is different).

I have confirmed that the evaluate results match the results from before (except they fix the cross-join case).

Changes

1. set default columns in new/empty DataFrames to avoid key error during subsequent joins.
2. use description field in all joins during evaluate.

Test

Following CMDs have been tested:

Internal Usage:

python qualx_main.py evaluate

[parthosa]

Thanks @leewyang. LGTME.

[parthosa]

In the python tool, we use a config file to store constants like these. It can be loaded using the class YAMLPropertiesContainer. In future, we could think of having a similar config file for QualX.

Example,

spark-rapids-tools/user_tools/src/spark_rapids_pytools/resources/qualification-conf.yaml

Lines 20 to 27 in b30ca6d

	summaryReport:
	fileName: rapids_4_spark_qualification_output.csv
	columns:
	- Vendor
	- Driver Host
	- Cluster Id
	- Cluster Name
	- App Name