use LinearSolve precs setup

[oscardssmith]

This updates NonlinearSolve to use the new precs interface in LinearSolve.

[avik-pal]

This would require a major version bump.

[mpf]

See also this Julia Discourse post.

[ChrisRackauckas]

Is this needed?

[oscardssmith]

it's not needed, but will be nice (as it will bring the interface inline with LinearSolve and OrdinaryDiffEq@V7

[ChrisRackauckas]

okay, I thought it was covered by another PR already.

[oscardssmith]

#465 was a simpler fix that fixed the bug that led me to create this PR, but all that did was switch NonlinearSolve from a cursed diagonal preconditioner to a real diagonal preconditioner, whereas this PR removes the preconditioner from the NonlinearSolve side entirely since the LinearSolve side can do all the work.

[oscardssmith]

@avik-pal if I rebase this, do you think we're ready to do the version bump required? I think this is a good change, but I don't really want to go through the rebase until you think we're conceptually ready to merge.

[avik-pal]

not yet, we need #458 to be merged before doing a bump

[avik-pal]

@oscardssmith we can get this in now