fix microsoft/qlib#1893

[csh589]

• adapt pandas >= 1.5.0
• add group_keys when the result from apply is a like-indexed Series or DataFrame
• replace some reset_index(level=1, drop=True) operation to group_key=False for efficency
  https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.groupby.html

Description

Motivation and Context

How Has This Been Tested?

• Pass the test by running: pytest qlib/tests/test_all_pipeline.py under upper directory of qlib.
• If you are adding a new feature, test on your own test scripts.

Screenshots of Test Results (if appropriate):

1. Pipeline test:
2. Your own tests:

Types of changes

• Fix bugs
• Add new feature
• Update documentation

[csh589]

@microsoft-github-policy-service agree