It produces an unrealistic distribution of topoindex.

Now the sum of the topoindex is done correctly along the graph and then averaged over all branches. This is done for the first HTUs which are selected in routing_reg_findbasin.

Some clean-up in the options to compute topoindex per HTU. A 4th option was introduced to take the mean of topoindex along the river within the HTU.

Add the distribution of the time step limitation for all HTUs. Serves as a diagnostic for the most approriate time step.

Improves the treatment of "undefined HTUs". There was a lack of consistency between the Python and F90 part. This is now clarified and the routing-graph.nc file can more easily be processed.

Attempts to add up-stream error into the grouph file. But the methodology is not good enough. It uses too much memory. The merger between the procs needs to be optimized.

Adds a function to gather and merge lists onto roo proc. It works based on a list of indicis already assembled on the root proc and then add to this list other information which is distributed over other procs.