performance improvement - split radiation into slow and fast routines

[rgknox]

From an algorithmic perspective, radiation scattering in the canopy is governed by slow (plant canopy composition and structure) and fast (incident radiation intensity and zenith angle) processes.

subroutine PatchNormanRadiation is where we handle scattering. This routine performs calculations that rely on both fast and slow.

Speaking generally, I think there is a way to split this routine so that we can preprocess as much of the slow timescale stuff as possible, once a day, and therefore reduce the number of computations that happen at sub-hourly calls (zenith stuff). I'm not exactly sure how much things can be re-organized, but I think its worth a hard look.

[rosiealice]

I agree @rgknox. In general it could obviously do with an overhaul and there's lots that could be pulled out here. Plus there is still, I think, an issue with radiation conservation in non SP modes, and that might help improve things.