Solver improvements

[jcitrin]

• Restrict Newton-Raphson exploration range for x_new, avoiding residual calls if temp or density < 0
• Introduce scalings on state or residual blocks such that all equations are of order 1
• Change residual_scalar in newton_raphson_solve_block to be defined by a 2-norm

[harshtech123]

@jcitrin i loves to bring this improvements into the picture can you please assign this to me?

[jcitrin]

Evaluating the impact of the changes may require some domain expertise, since the sim integration tests results will slightly change for the 2nd and 3rd tasks here, and it's important to have some experience with the simulations to judge if those changes are OK or not, as well as reevaluate the tests etc. (see https://torax.readthedocs.io/en/latest/contribution_tips.html#contribution-tips)

Perhaps try going ahead with the first task, which should not break any tests and should make a slight speedup for Newton-Raphson cases that can be evaluated.

To set expectations, please be aware we cannot guarantee to iterate on the eventual PR if it looks like the review process will be very time-consuming. Apologies for that but we have limited bandwidth.

[harshtech123]

thank you ! i'll look closely before making any changes .