Return evaluation results to callers

[tleyden]

In order for callers to use the evaluation results, for example displaying them in a UI, the evaluation functions should return these results to the caller.

This PR:

1. Defines a container class for the eval results
2. Decouples the calculation and printing of eval results
3. Returns eval results as a container class instance

It uses Pydantic for the container class in case we later need to use validation or any other features offered by pydantic. It requires the same pydantic version as used in the Arcee platform.

Testing

Manually tested, here is output:

10/25/2023 10:53:29 - INFO - dalm.eval.eval_retriever_only - Construct passage index
10/25/2023 10:53:29 - INFO - dalm.eval.eval_retriever_only - Evaluation start
10/25/2023 10:53:30 - INFO - dalm.eval.utils - Retriever results:
10/25/2023 10:53:30 - INFO - dalm.eval.utils - Recall: 0.883495145631068
10/25/2023 10:53:30 - INFO - dalm.eval.utils - Precision: 0.08834951456310675
10/25/2023 10:53:30 - INFO - dalm.eval.utils - Hit Rate: 0.883495145631068
10/25/2023 10:53:30 - INFO - dalm.eval.utils - *************
10/25/2023 10:53:30 - INFO - root - Retriever evaluation results: total_examples=206 recall=0.883495145631068 precision=0.08834951456310675 hit_rate=0.883495145631068