Stream Flow Routing (SFR)

I took a look at the SFR package in the weekend, a few notes:

• For connection data (from reach to reach), an xugrid method was still missing - I've added directed_edge_to_edge_connectivity to xugrid for this (still unreleased).
• The latest version of MODFLOW 6 appears to have new SFR features in the form of diversions and profiles (storage profiles), but these aren't well documented yet
• For diversions and profiles, it could be useful to create helper classes similar to those used for lakes
• But for now, we'll leave that out since I don't think the table formats are even properly documented(?)
• Furthermore, the period data has a very peculiar structure, where updates are specified per parameter

A package with basic functions is relatively straightforward: you use the edge centroids to find in which cell a reach is located, and then you write all the input (just like Well).

A bit more logic is required to do assignment to any grid; the network needs to be intersected with a (2D) grid.
This isn't that hard to do, numba_celltree already has the intersection logic. I.e. a regrid method might not work; but I don't think this is strictly necessary: if the user's network topology is fine enough, it'll go just fine. (In case of a coarse network, the exchange with the groundwater just ends up being very "patchy" and physically non-sensical -- a surface water every 5th groundwater cell or so -- but the model should run fine.)

Start by reading the existing SFR package and the Well implementation referenced in the issue, then review how xugrid connectivity and numba_celltree grid intersections could be used. The scope is not settled: the basic input-writing functions, arbitrary-grid assignment, and possible diversion or profile helpers are discussed but not selected, so completion criteria need clarification.