Add GPURodas3 and GPURodas42 support to EnsembleGPUKernel

[AJ0070]

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

This PR introduces GPU kernel solver support for Rodas3 and Rodas42 in DiffEqGPU, following the existing Rosenbrock-style implementation pattern.

Key changes

• Added new implicit GPU algorithm types: GPURodas3 and GPURodas42
• Implemented method metadata and order wiring for both algorithms
• Added Rodas tableaus and coefficients required for kernel execution
• Updated stiff integrator type and initialization paths for Rodas3 and Rodas42
• Implemented kernel perform-step logic for Rodas3, along with module includes and exports
• Expanded test coverage across GPU kernel tests, lower-level API tests, and JET checks

AI Disclosure: used copilot for understanding previous code and generating parts of code(tested and verified manually)

[oscardssmith]

At some point we should unify all the rosenbrocks (like we've done on the OrdinaryDiffEq side), but this looks reasonable.

[ChrisRackauckas]

I really don't see the point of this.

[oscardssmith]

George Gkountouras requested it (presumably to have a low order rosenbrock method that works for DAE GPUs).