Azure ML - AIO implementation of OBO credential uses an undefined class attribute causing token acquiral to fail

- **Package Name**: azure-ai-ml
- **Package Version**: 1.26.0
- **Operating System**: Ubuntu
- **Python Version**: python 3.11

**Describe the bug**
[_identity_config is not defined anywhere in the parent/child classes](https://github.com/Azure/azure-sdk-for-python/blob/1f7312f040269e1d0a31d463f93b0251c80cd4c2/sdk/ml/azure-ai-ml/azure/ai/ml/identity/_aio/_internal/managed_identity_client.py#L31)

**To Reproduce**
Steps to reproduce the behavior:
1. Instantiate the `AzureMLOnBehalfOfCredential` class
2. Attempt to retrieve a token as part of an aio pipeline
3. Raises `AttributeError: object 'AsyncManagedIdentityClient' has no attribute '_identity_config'

**Expected behavior**
Acquiring an OBO token via an aio event loop is possible. As it stands the OBO AIO credential class cannot be used to perform any read/write operations that leverage an fsspec implementation (since this uses aio).