Deriving the outlet and area of waterbodies

[hboisgon]

The method that is used to derive waterbodies outlet and area should be improved as it can lead to some wrongful delineation.
So far, we use:

1. Rasterize waterbody polygon using all_touched = True
2. Derive the outlet using the cell within the rasterized polygon with max upstream area value

Problem is that with this technique we do not really follow the flow directions so we can include cells in the mask that are part of a different subbasin/tributary. More problematic, in some cases the outlet of the waterbody falls in another basin. Example for the Danube near the sea outlet where a small lake is too close to Danube and the method above then puts the outlet of the lake in the Danube itself (I think this also happens for the Nile in the wetlands part):

Not too sure how to improve here but maybe all touched is a little radical to start with so maybe we should exclude cells from the area if the coverage of the cell by the lake is less than a certain threshold like 10%? (there's a nice raster.rasterize_geometry method in core to do that now too). Or used a different rasterization method than all_touched is True?
For the outlet, we should still make sure it falls under the right basin. The 10% filter might already help in some cases but we could also use the watershed area (Wshd_area) ie upstream area column of the waterbody database if they have it for better snapping. This type of snapping is also under dev in PR #172 for gauges.