This adds the mean, minimum, and maximum padding modes as aggregations. Additionally, it also contains code for a median mode, but this is broken because median does not support the where parameter. Fixed by moving to nanreduce, which also allows dask to work properly.
These modes are implemented by searching the neighbours of the padded cells. Depending on how costly this is, we might want to switch to grouping the neighbours of the input cells by the cell ids of the padded cells.
This adds the
mean,minimum, andmaximumpadding modes as aggregations.Additionally, it also contains code for aFixed by moving tomedianmode, but this is broken becausemediandoes not support thewhereparameter.nanreduce, which also allowsdaskto work properly.These modes are implemented by searching the neighbours of the padded cells. Depending on how costly this is, we might want to switch to grouping the neighbours of the input cells by the cell ids of the padded cells.