Propagation of nodata in routing modules

[davemfish]

A nodata pixel in an input dataset (e.g. precipitation, erosivity, etc) that occurs in a location with valid DEM data can result in a large swaths of nodata in the results of a routing function. This issue is well-described in this SDR example: #1433

I'm creating a new issue because it has also recently come up in SWY, and could possibly occur in other models that include a routing algorithm. https://community.naturalcapitalproject.org/t/base-flow-and-aggregate-results-are-coming-with-nodata/5597

I think the large swaths of nodata is the "correct" result of these algorithms, but it's confusing to users and time-consuming to trace back to the original raster with the missing data. With that in mind, it might make sense to apply the solution that James used in SDR to the other models where this can happen.

I think the result would be very small, isolated occurrences of nodata, instead of large swaths. And it would be easier to trace the nodata in the outputs to nodata in the inputs, since the nodata would not be able to "walk" very far away from its source.