Smoothing of river parameters

[hboisgon]

I did some checks on the the river parameters we estimate for river slope / width /depth and found some issues and/or wishes.

1. For river slope, we can decide to compute slope over a different length with slope_len but actually smoothing is not yet possible. Would be nice to have, as in kinematic wave, water levels and volumes are very very strongly dependent on river slope values. Would also be nice to add a minimum value for slope, as the default one is zero and the one in Wflow.jl can sometimes be too low.
2. River width and depth: there should be smoothing, but I actually found a bug (should be max instead of min):

hydromt_wflow/hydromt_wflow/workflows/river.py

Lines 269 to 271 in 5e59664

	if smooth_len > 0:
	nsmooth = min(1, int(round(smooth_len / rivlen_avg / 2)))
	kwgs = dict(n=nsmooth, restrict_strord=True)

3. To prepare width and depth, we can use values from a river geometry file (like Lin or HydroAtlas). However, in these databases, smaller rivers are very often not included and it can happen that along a river stretch, some portions will have values, then no data, then values again. Or currently, when filling nodata gaps, we use a constant value (eg 1 meter), so the case can happen that along a river, the depth can be around 3 meters then 1 meter in the nodata gap, then back to 3-4 meters when we have data again... Might be nice if possible to add a sort of interpolation method to fill in these nodata gaps (especially on qbankfull which is used for computing depth).