NDR runoff proxy index depends on watershed extent

[emlys]

@phawthorne raised this issue:
The runoff proxy index is calculated as $RPI_i = RP_i / RP_{av}$ (dividing runoff proxy by average runoff proxy within the watersheds). This is problematic because average runoff proxy $RP_{av}$ changes depending on which pixels/watersheds are included in the raster. @phawthorne said they have even considered inserting fake pixel values into the RP raster in order to change the $RP_{av}$. We could expose the $RP_{av}$ parameter and allow the user to specify it instead.

[phawthorne]

Thanks for opening this, @emlys! An additional issue is that this makes it hard to run the model under different climate scenarios because any shifts in average precipitation between the baseline and future scenarios get normalized out so we can't actually assess the impact of the change. With an exposed $RP_{av}$ value, we could, for example, calculate $RP_{av}^{baseline}$ and then apply the same value to future scenarios.