OmegaConfigLoader's __repr__ is not implemented correctly, and it does not behave similar to a UserDict

[noklam]

Description

Currently, base_env and default_run_env is not included in the __repr__.

When it is printed, it shows something like this

In [11]: c
Out[11]: OmegaConfigLoader(conf_source=/Users/Nok_Lam_Chan/dev/kedro-inspect/demo-project/conf, env=None, config_patterns={'catalog': ['catalog*', 'catalog*/', '/catalog*'], 'parameters': ['parameters*', 'parameters*/', '/parameters*'], 'credentials': ['credentials*', 'credentials*/', '/credentials*'], 'globals': ['globals.yml']})

Context

Even it inherits from UserDict, but the dictionary interface seems broken. For example:

for key in config_loader.keys():
  print(key)

will print nothing even though config_loader["parameters"] exist. For the context, I was trying to build a plugin that iterate the config loader and "compile" the configuration.

I need to use a workaround looks like this:

keys = config_loader.config_patterns.keys()
for key in keys:
   config = config_loader[key] 
   ...

In addition, if __repr__ is removed, the config_loader will be printed as if it is a "dictionary". Right now it always print the __repr__ regardless of the keys, values, items or print method.

Cc @merelcht