fix: parse mixed Yahoo date index

[he-yufeng]

What

• make Yahoo normalization tolerate a mixed date index containing both daily dates and timezone-bearing timestamps
• keep the existing timezone-stripping behavior after parsing
• add a small regression test for the mixed daily/timestamp shape from 执行yahoo normalize data 发生 ValueError: unconverted data remains when parsing with format "%Y-%m-%d" 错误 #2014

To verify

• python -m py_compile scripts\data_collector\yahoo\collector.py tests\test_yahoo_collector.py
• git diff --check -- scripts\data_collector\yahoo\collector.py tests\test_yahoo_collector.py

Local note

• python -m pytest tests\test_yahoo_collector.py -q is blocked in this Windows checkout because qlib's local C extensions are not built: ModuleNotFoundError: No module named 'qlib.data._libs.rolling'.
• python setup.py build_ext --inplace also fails before tests with Conda Python's pyconfig.h unable to include io.h, so I did not classify that as a patch failure.

Fixes #2014