Apply Dask to parallelise acs_regional_stats

[xenct]

acs_regional_stats can be very memory intensive to run, particularly over many regions and many timesteps.
We should develop an example of running acs_regional_stats for many years of daily data to produce area averaged timeseries for regions. Currently, this is possible, but will take several minutes to calculate.
Dask is likely to be able to achieve this by calculating area averages per file.
Previous development has focused on reducing memory usage through other clever means, such as implementing chunks to reduce the number of timesteps loaded into the memory to calculate stats over each time. This could be parallelised, but it is not currently.