fix: handle string date indexes in gap stats

[ikejourney]

Summary

• handle string date indexes when computing time series gap stats
• convert parseable string index values to datetimes before calling diff()
• return empty gap stats for non-numeric, non-datetime, non-parseable indexes instead of raising
• add a regression test for string date indexes

Closes #1773.

Checks

• python3 -m py_compile src/data_profiling/model/pandas/describe_timeseries_pandas.py tests/unit/test_time_series.py
• git diff --check

I could not run the pytest target locally in this Codex environment because pytest and the project runtime dependencies, including pandas, are not installed.