[Bug] Join failed to get groupBy module on different file structure

[mickjermsurawong-openai]

• Join code expects to get module object from the groupby that looks like below
  

  chronon/api/py/ai/chronon/join.py

  Lines 65 to 67 in a3983fb

  	for ref in gc.get_referrers(group_by):
  	if "__name__" in ref and ref["__name__"].startswith("group_bys"):
  	group_by_module_name = ref["__name__"]

{'__name__': 'chronon_features.joins.quickstart.training_set', 
'__doc__': None, 
'__package__': 'chronon_features.joins.quickstart', 
'__loader__': <ddtrace.internal.module._ImportHookChainedLoader object at 0x110e7aed0>, '__spec__': ModuleSpec(name='chronon_features.joins.quickstart.training_set'
...
}

• One potential problem, when referrer of the group_by isn't a dict or iterable, (In my case JoinPart object itself is the referrer, this check would fail

TypeError: argument of type 'JoinPart' is not iterable

• Second potential problem, the check of name starting "group_bys" assumes that we have group_by as top-level. In my case, i have a separate python project chronon_features, so it would expect module name as chronon_features.group_bys

[nikhilsimha]

chronon is very convention driven & expects the following directory structure

your_top_level_directory # i recommend calling it chronon
   /joins
      /team1
         join1.py
         join2.py
         .
         .
      /team2
      .
      .  
   /group_bys
      /team1
         groupBy1.py
         .
         .
     /team2
     .
     .
   /staging_queries 

[mickjermsurawong-openai]

hey @nikhilsimha thanks for taking a look, i made a repro to motivate my use case, pls help to take a look. not urgent
#868