Illegal Comparisons when handling pandas DataFrame objects

[jeffknaide]

Describe the bug
In several places, the code attempts to check if a node is missing with a comparison + a conditional (eg x = "a" if node == "???" else "b"). This conditional breaks if the node value is a pandas DataFrame, as DataFrames implement additional __bool__ logic that prevents such a conditional (see: https://github.com/pandas-dev/pandas/blob/v1.2.2/pandas/core/generic.py#L1440-L1446).

To Reproduce
creating a DictConfig:

from pandas import DataFrame
from omegaconf import OmegaConf

kwargs = OmegaConf.create({"a": DataFrame()}, flags={"allow_objects": True})

when attempting to set an attribute of an existing config:

from pandas import DataFrame
from omegaconf import OmegaConf

c = OmegaConf.create()
c._set_flag("allow_objects", True)
c.a = DataFrame()

Expected behavior
when running creating DictConfig or setting a new attr, there should simply be an update to the config where the DF object is now included (assuming allow_object flag is set to True).

Additional context

• OmegaConf version: master branch
• Python version: 3.7.6
• Operating system : ubuntu
• Please provide a minimal repro