Convert litter model to use biomass triplets

The litter model is better off tracking masses of nitrogen and phosphorus, rather than the stochiometric ratios, because averaging stoichiometric ratios is incredibly easy to screw up. If this was the case pools could be expressed as biomass triplets, which would also simplify the description of the model quite a lot.

This would be a large but fairly straightforward refactor of the litter model