Implement Season Period Object

[paololucchino]

Filtering and aggregating timeseries by season is a common operational step in our work.

Noting the usefulness of the Dekad object and associated accessor, this issue proposes the creation of a Season object and associated accessor.

Example functionality would be something like the below. An added complication compared to Dekad is that the user would need to feed in some start and end range.

Example time dimension
[2020-01-11, ...., 2020-03-21, 2020-04-01, ...., 2021-02-11, ..., 2021-03-01]

# Pseudo code converting time dimension to season dimention
# Must pass season_range in Dekad of Year format (0...35)
# season_range could be a list of tuples to handle areas with more than one season
da.time.season.label(season_range=(5,16))

Result in format : YYYY_SeasonNumber
[NaT, ...., 2020_1, 2020_1, ....,NaT, ..., 2021_01]

This new index would then easily allow to drop out of season timesteps and run groupby operations by season.