`disaggregate` does not preserve extent

[tiemvanderdeure]

Now that it's not possible to pass a locus to disaggregate, it always disaggregates at the Start, which returns some pretty unexpected dimensions (to me). I think it should be a priority to always return the same bounds when disaggregating - right now that doesn't happen.

using Rasters, Rasters.Lookups
ras = Raster(rand(X(Sampled(5:10:95; sampling = Intervals(Center())))))
ras2 = disaggregate(ras, 10)

extent(ras2) == extent(ras)

I thought this would be an easy fix, but it requires changing some of the internals and then everything gets messed up.

So maybe the solution is to shiftlocus internally or something. Or do should we make it possible to pass a locus again?

[rafaqz]

Shouldn't it use the existing locus? Maybe I messed that up. (also please print outputs of these example ;)

[tiemvanderdeure]

No it just uses Start() - but even if it uses the existing locus it doesn't always work

[tiemvanderdeure]

Okay I realised this also affects aggregate. I know this might be trickier than for disaggregate but I think it's worth a bit of effort to get this right

x = X(Sampled(5:10:95; sampling = Intervals(Center()))) |> Rasters.format
xag = aggregate(sum, x, 2)

bounds(x) # (0.0, 100.0)
bounds(xag) # (5.0, 105.0)

[rafaqz]

Can you write up a PR with some failing tests with exactly the behaviour you want? I can make it work, but I'm not sure I totally understand the context