Chunking changed for function over time axis

The following function seems to destroy the chunking of my data when it is used on a larger dataset.
It keeps the chunking, when it is applied on a small testset.

function cube_eemd(xout, xin)
    # libeemd segfaults when using Float32 data
    ind = .!ismissing.(xin)
    ts = Float64.(collect(xin[ind]))
    xout[ind] .= ceemdan(ts, 3)[:, end]
    xout[.!ind] .=missing
end

function eemdfilter(cube)
    indims = InDims("Time")
    outdims = OutDims("Time")

    efcube = mapCube(cube_eemd, cube, indims=indims, outdims=outdims, max_cache=1e8)

    newcube = mapCube(copyto!, efcube, indims=InDims("Lon", "Lat"), outdims=OutDims("Lon", "Lat"), max_cache=1e8)

    return newcube
end

This was reported as https://github.com/esa-esdl/ESDL.jl/issues/127.

Start by reproducing the larger-dataset case in the eemdfilter function, focusing on the two mapCube calls and the cube_eemd callback. Compare chunking before and after applying the function, and use the small test set as a baseline. Done means the function preserves chunking for larger datasets as well as small ones.