make a history dimension for hydraulic soil radial dimension

[ckoven]

Given that the soil model includes a radial dimension around roots in each soil layer which it represents as multiple soil shell layers, it would be good to allow the model to properly output that. Right now this is handled via a bit of a work-around hack, e.g. in these places:

https://github.com/NGEET/fates/blob/master/main/FatesHistoryInterfaceMod.F90#L387
https://github.com/NGEET/fates/blob/master/main/FatesHistoryInterfaceMod.F90#L2906
https://github.com/NGEET/fates/blob/master/main/FatesHistoryInterfaceMod.F90#L3166

This is fragile since the size x age class dimension could be smaller than the nlevsoi x nshell, which would likely give buffer overflows. So I think we should pull this dimension out and make it a thing.

[ckoven]

also related, the soil water uptake diagnostic is using separate variables for each layer instead of using an nlevsoi-dimensioned variable. would be good to consolidate and dimension properly, both the history variable and the cohort-level variables tracking these diagnostic quantities. e.g. these places:
https://github.com/NGEET/fates/blob/master/biogeophys/FatesPlantHydraulicsMod.F90#L2844
https://github.com/NGEET/fates/blob/master/main/FatesHistoryInterfaceMod.F90#L3065
https://github.com/NGEET/fates/blob/master/main/FatesHydraulicsMemMod.F90#L269
Sorry to be nitpicky, but this all threw me off while trying to read it and work backwards from the history outputting...

[rgknox]

This diagnostic should help us make sure that #504 is no longer an issue