Update the default theta parameter definition in LM & LMTR and remove `νInv`

[MaxenceGollier]

Currently, we compute the steps in LM and LMTR as νInv = (1 + θ) * (σmax^2 + σk) in LM and νInv = (1 + θ) * σmax^2 in LMTR.
It would be more readable (and actually more compliant with the article notation to have respectively νInv = (σmax^2 + σk)/θ and νInv = (σmax^2/θ instead.

I will open a PR to change the default value of θ accordingly and do these modifications. @dpo It is related to a comment you made in #172 .
@MohamedLaghdafHABIBOULLAH fyi.

[dpo]

I think that’s a good idea, but, in addition, I also wish we removed νInv completely, and just used ν everywhere instead.

Also, the choice of ν in LMTR should involve Δ as it does in TR.