Rethinking litter decay models

[hrlai]

While addressing #101 I estimated equilibrium root structural stock that is incredibly high. It was probably due to very slow decaying rate of this litter component.

I then revisited the litter decay model in the merged PR #70 . On a closer look, I have a gut feeling that some of the parameters are "off"; it's not that they are wrong, it's just that they might have been strongly influenced by my prior choice. One of them is they key parameters $r$, which determines the effect of lignin on decay rates. It's really close to my prior, and I suspect that it has a lot to do with weak data going into the model --- lignin data is really crappy for the leaf and wood decomposition SAFE datasets I used.

In short, I think lignin is the bottleneck for us right now.

I then tried to find datasets from scratch again, this time manually and using LLMs. No good yield. There simply wasn't enough studies that collected mass loss over time and lignin per sample.

Next step, I will scrutinise our model in #60 again, and think of ways to save parameters. My gut feeling is that some of them should be reparameterised together and then recovered post-fiiting with (lots of) assumptions, unfortunately.

[hrlai]

Note to future self: if we ever revisit these models, consider a QR decomposition since C:N and C:P (and lignin) are all likely to be positively correlated.