Better SOI series

[CommonClimate]

When working on PR #662 , I discovered that the original soi_data.csv had a wonky fractional year time axis that got weirdly parsed by tbase.time_to_datetime(), resulting in a DatetimeIndex with no month of February, and 2 months of March. The series was also fairly short. I downloaded a new csv file from NOAA PSL and (with Claude Code's help) amended the load_dataset() logic to parse dates if they are available (as in this case), resulting in a proper, monthly time axis. The new code should be robust for future additions, and ensures that SOI is a well-behaved example for what we need it to show.

[khider]

Docs are not building however:

sphinx.errors.ExtensionError: Cell raised uncaught exception:
�[31m---------------------------------------------------------------------------�[39m
�[31mNameError�[39m Traceback (most recent call last)
�[36mCell�[39m�[36m �[39m�[32mIn[5]�[39m�[32m, line 11�[39m
�[32m 7�[39m ts = pd.Series(values, index=dates, name=�[33m'�[39m�[33mmonthly_values�[39m�[33m'�[39m)
�[32m 9�[39m �[38;5;66;03m# Example 1: Single year (int)�[39;00m
�[32m 10�[39m �[38;5;66;03m# Compute Jan-Mar average for 2021 only�[39;00m
�[32m---> �[39m�[32m11�[39m avg_single = �[43mcustom_year_averages�[49m(ts, �[32m1�[39m, �[32m3�[39m, years=�[32m2021�[39m)
�[32m 12�[39m �[38;5;28mprint�[39m(�[33m"�[39m�[33mSingle year (2021):�[39m�[33m"�[39m)
�[32m 13�[39m �[38;5;28mprint�[39m(avg_single)

�[31mNameError�[39m: name 'custom_year_averages' is not defined

Extension error:
Cell raised uncaught exception:
�[31m---------------------------------------------------------------------------�[39m
�[31mNameError�[39m Traceback (most recent call last)
�[36mCell�[39m�[36m �[39m�[32mIn[5]�[39m�[32m, line 11�[39m
�[32m 7�[39m ts = pd.Series(values, index=dates, name=�[33m'�[39m�[33mmonthly_values�[39m�[33m'�[39m)
�[32m 9�[39m �[38;5;66;03m# Example 1: Single year (int)�[39;00m
�[32m 10�[39m �[38;5;66;03m# Compute Jan-Mar average for 2021 only�[39;00m
�[32m---> �[39m�[32m11�[39m avg_single = �[43mcustom_year_averages�[49m(ts, �[32m1�[39m, �[32m3�[39m, years=�[32m2021�[39m)
�[32m 12�[39m �[38;5;28mprint�[39m(�[33m"�[39m�[33mSingle year (2021):�[39m�[33m"�[39m)
�[32m 13�[39m �[38;5;28mprint�[39m(avg_single)

�[31mNameError�[39m: name 'custom_year_averages' is not defined

Not sure If the error is due to this PR or a merge from last week that wasn't checked properly

[khider]

Problem with the docs. See comment in the PR