Add static StrongWolfe path for OptimizationProblem

[AdityaPandeyCN]

Checklist

• Appropriate tests were added
• Any code changes were done in a way that does not break public API
• All documentation related to code changes were updated
• The new code follows the
  contributor guidelines, in particular the SciML Style Guide and
  COLPRAC.
• Any new documentation only uses public API

Additional context

StrongWolfeLineSearch's static (GPU/SArray/Number) path hardcoded the residual merit ½‖f(u)‖². That is correct for NonlinearProblem, but wrong for minimization: an objective line search needs to test decrease of f itself.
Passing an OptimizationProblem objective through the existing path would silently evaluate ½‖f‖² instead.
This selects the merit by problem type on the static path:

• AbstractNonlinearProblem --½‖f(u)‖² (unchanged)
• OptimizationProblem -- f(u) (static-only; non-static u/fu raise ArgumentError)
  Dispatch uses zero-sized mode structs in the cache so the search loop stays branch-free / warp-uniform. grad_f must match the merit: J'f for residual mode, ∇f for scalar mode. Static solve! also accepts a per-call α_max (clamped by the cache); on failure it still returns α_init.

[AdityaPandeyCN]

@ChrisRackauckas Can you take a look?