Framework/tool for tracking PyTorch UT failures

[albmalamd]

Propose some tools/framework for recording/assigning/tracking failures seen across different branches/architectures/ROCm versions

[albmalamd]

Had a conversation with Jithun and Michael and following steps will be taken by me:

• create a demo with django for database manipulation (sqlite3 as starting point)
• implement a function for translating django output to pandas.DataFrame