Add support for ODE_DEFAULT_NORM for AbstractRaggedVectorOfArray

[JoshuaLampert]

This adds support for UNITLESS_ABS2, recursive_length, and ODE_DEFAULT_NORM for AbstractRaggedVectorOfArray from RecursiveArrayToolsRaggedArrays.jl as discussed in SciML/RecursiveArrayTools.jl#582 (comment).

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

Add any other context about the problem here.

[ChrisRackauckas]

Again this introduces a circular, it cannot go here.

[ChrisRackauckas]

I'll take it from here.

[ChrisRackauckas]

Wait, I'm so dumb.

[ChrisRackauckas-Claude]

Diagnosis for test (OrdinaryDiffEqNonlinearSolve_ModelingToolkit, 1, ubuntu-latest, 120, 1) failure on this PR (job 73389128637):

Failure is pre-existing and unrelated to this PR. The 4 test failures all come from lib/OrdinaryDiffEqNonlinearSolve/test/modelingtoolkit/dae_initialize_integration.jl:98-101, in the @testset "\reinit!` updates initial parameters"block. Afterreinit!(integ, new_u0)withnew_u0corresponding tox => 2.0, y => 3.0, the Initial(x)/Initial(y)parameters in the integrator (and the integrator's symbolic stateinteg[x], integ[y]) remain at their original values (1.0, 2.0) instead of being updated to (2.0, 3.0). The system is a standard ODE solved with Tsit5(), so no implicit-method norm is invoked and AbstractRaggedVectorOfArray` is not in play anywhere.

Evidence it is pre-existing:

• This PR's diff only adds three method dispatches in lib/DiffEqBase/src/common_defaults.jl (UNITLESS_ABS2, recursive_length, ODE_DEFAULT_NORM for AbstractRaggedVectorOfArray) and a deps/test entry — none of these can affect MTK reinit! behavior on a non-ragged ODEProblem.
• The exact same 4 test failures (dae_initialize_integration.jl:98-101) appeared on PR branch fix-qmin-controller-access (job 73293875132) on 2026-04-27, before this PR existed.

Recommendation: this failure should be tracked separately — likely an MTK / SymbolicIndexingInterface regression where reinit! no longer propagates the new u0 into the Initial(...) parameter mirror. Not blocking for this PR (which has already been merged in 3a6e660).

Co-Authored-By: Chris Rackauckas accounts@chrisrackauckas.com