Add option for longer dynamic spinups

[pat-schmitt]

In this PR, I introduce a new methodology that allows for longer spinup periods during dynamic spinup, particularly when creating the preprocessed directories. When dynamic spinup was initially developed, historical data was only available from 1979, so the spinup in the prepro levels was limited to that year. However, historical data is now available from 1901, and recent studies have shown that some glaciers require a longer spinup period.

Main changes:

• New kwarg spinup_periods_to_try in tasks.run_dynamic_spinup: accepts a list of additional spinup periods, which are tried in order if both the initially defined spinup period (spinup_period_initial) and the minimum spinup period (min_spinup_period) fail. This means we first try the old default workflow, and only if this fails we try the additional spinup periods.
• New argument --dynamic-spinup-periods-to-try in prepro_levels.py, defaulting to [30, 40, 50, 60, 70, 80, 90, 100]. To disable this new behaviour, set --dynamic-spinup-periods-to-try none.
• No additional tests were added. The change was only tested by running RGI region 19 (Antarctica).
• The resulting model output _historical_spinup in the prepro levels still starts from 1979, which is achieved via fixed_geometry_spinup_yr. If fixed_geometry_spinup_yr (e.g. 1979) is later than FlowlineModel.yr (e.g. 1950), the model runs until fixed_geometry_spinup_yr before storing any output. This works for now, but should eventually be made more explicit. For example, by adding a new kwarg such as start_yr_to_store_model_output.

Related to #1837: but no real optimization to find the best spinup time, we just look for any working spinup time, and try to have it as short as possible, up to a maximum of 100 years spinup

• Tests added/passed
• Fully documented
• Entry in whats-new.rst