Describe the bug
isinstance(config_dict_obj, collections.abc.Mapping) returns False even though it is a "dictionary" type

This is because class ml_collections.config_dict.ConfigDict does not inherit Python built-in abstract base class collections.abc.Mapping, despite being a "dictionary" type, producing unexpected behavior when using reflection or duck-typing with isinstance(x, collections.abc.Mapping). This is especially problematic for serializing and logging code where it is common to apply different formatting to different collection types.

See PR #46 for proposed solution.

To Reproduce
The following code snippet reproduces and illustrates the bug

import collections.abc
from typing import Any
from ml_collections import config_dict


def safe_print(obj: Any):
    """Print object with masked values to avoid leaking API keys in the logs"""
    if isinstance(obj, collections.abc.Mapping):
        for k, v in obj.items():
            print(f"{k}={'*' * len(str(v))}")
    elif isinstance(obj, collections.abc.Sequence):
        print(f"[ ... ({len(obj)} items)]")
    else:
        raise ValueError(f"Unsupported type: {type(obj)}")


safe_print({"a": 1})
>>> a=*

conf = config_dict.ConfigDict({"a": 1})
safe_print(conf)
>>> Traceback (most recent call last):
>>>   File "<stdin>", line 1, in <module>
>>>   File "<stdin>", line 8, in safe_print
>>> ValueError: Unsupported type: <class 'ml_collections.config_dict.config_dict.ConfigDict'>

Expected behavior
isinstance(config_dict_obj, collections.abc.Mapping) should return True

Environment:

• OS: Linux (Pop_OS/Ubuntu)
• OS Version: 6.9.3/22.04
• Python: Python 3.10

Additional context
Correctly inheriting collections.abc.Mapping would also improve Python static type analysis tooling (eg Ruff)