support for padding nodata in timeseries

[jdries]

Support something like:

def fill(array):
   first = first(array,ignore_nodata=true)
   last = last(array,ignore_nodata=true)
   frontfilled = array_apply(array,if(nodata(x)&& index<middle,first,x)
   fully_filled=array_apply(frontfilled, if(nodata(x)&& index>=middle,last,x)
   return fully_filled
apply_dimension(dim='t',callback=fill)

The goal is to created a timeseries without nodata, as some methods can not deal with that. Linear interpolation does not extrapolate, so typically can return timeseries with nodata at the start or end.

[soxofaan]

FYI related discussion

• array_interpolate_linear: toggle to also extrapolate leading/trailing no-data openeo-processes#510