Widen DiffEqBase/SciMLBase/RecursiveArrayTools compat to admit v7/v3/v4

[ChrisRackauckas-Claude]

Summary

The SciML ODE stack just shipped a coordinated breaking release: OrdinaryDiffEq.jl v7, DiffEqBase v7 (which requires SciMLBase v3), SciMLBase v3, and RecursiveArrayTools v4. BoundaryValueDiffEq's [compat] currently caps all three at the older majors, so any downstream project that depends on both BVDE and OrdinaryDiffEq v7 fails to resolve. Concretely, this was hit by the OrdinaryDiffEq monorepo's DiffEqDevTools test env — BVDE 5.21.0's DiffEqBase "6.183" pin blocks the DiffEqDevTools test env from resolving.

This widens the per-package [compat] entries to admit the new majors (union style so existing consumers are not broken):

• DiffEqBase: "6.183" → "6.183, 7" (top-level), "6.213" → "6.213, 7" (Core)
• SciMLBase: "2.152.1" → "2.152.1, 3"
• RecursiveArrayTools: "3.27.0" / "3.31.2" → "..., 4"

Bumps the affected package versions (BoundaryValueDiffEq 5.21.0→5.22.0, Core 2.2.0→2.3.0, Ascher/MIRKN 1.12.0→1.13.0, MIRK/FIRK 1.13.1→1.14.0, Shooting 1.13.0→1.14.0).

No source changes: qualified uses like SciMLBase.DEStats still resolve under v3. If CI surfaces v3/v4-specific issues they will need individual attention, but the compat widening alone unblocks the resolver.

Test plan

• Package tests pass on the old stack (SciMLBase 2.x / DiffEqBase 6.x / RAT 3.x)
• Package tests pass on the new stack (SciMLBase 3.x / DiffEqBase 7.x / RAT 4.x)
• Downstream OrdinaryDiffEq.jl DiffEqDevTools test env resolves once this is registered

🤖 Generated with Claude Code

[github-actions]

Benchmark Results

Click to check benchmark results

	master	eb49c07...	master / eb49c07...
Simple Pendulum/IIP/BoundaryValueDiffEqMIRK.MIRK2()	0.507 ± 0.0083 s	0.509 ± 0.0031 s	0.997 ± 0.017
Simple Pendulum/IIP/BoundaryValueDiffEqMIRK.MIRK3()	10.5 ± 0.8 ms	10.5 ± 0.71 ms	1 ± 0.1
Simple Pendulum/IIP/BoundaryValueDiffEqMIRK.MIRK4()	2.24 ± 0.38 ms	2.29 ± 0.4 ms	0.979 ± 0.24
Simple Pendulum/IIP/BoundaryValueDiffEqMIRK.MIRK5()	2.74 ± 0.48 ms	2.75 ± 0.5 ms	0.994 ± 0.25
Simple Pendulum/IIP/BoundaryValueDiffEqMIRK.MIRK6()	1.31 ± 0.27 ms	1.31 ± 0.28 ms	0.997 ± 0.3
Simple Pendulum/IIP/MultipleShooting(10, Tsit5; grid_coarsening = false)	1.49 ± 0.47 ms	1.53 ± 0.45 ms	0.973 ± 0.42
Simple Pendulum/IIP/MultipleShooting(10, Tsit5; grid_coarsening = true)	3.06 ± 0.84 ms	3.15 ± 0.81 ms	0.971 ± 0.36
Simple Pendulum/IIP/MultipleShooting(100, Tsit5; grid_coarsening = false)	0.0412 ± 0.011 s	0.0372 ± 0.0043 s	1.11 ± 0.33
Simple Pendulum/IIP/MultipleShooting(100, Tsit5; grid_coarsening = true)	0.0556 ± 0.0098 s	0.0569 ± 0.011 s	0.977 ± 0.25
Simple Pendulum/IIP/Shooting(Tsit5())	0.216 ± 0.063 ms	0.218 ± 0.069 ms	0.991 ± 0.42
Simple Pendulum/OOP/BoundaryValueDiffEqMIRK.MIRK2()	0.638 ± 0.022 s	0.665 ± 0.017 s	0.958 ± 0.041
Simple Pendulum/OOP/BoundaryValueDiffEqMIRK.MIRK3()	13.6 ± 6.1 ms	13.8 ± 7 ms	0.984 ± 0.67
Simple Pendulum/OOP/BoundaryValueDiffEqMIRK.MIRK4()	2.75 ± 0.22 ms	2.79 ± 0.24 ms	0.986 ± 0.12
Simple Pendulum/OOP/BoundaryValueDiffEqMIRK.MIRK5()	3.37 ± 0.27 ms	3.4 ± 0.31 ms	0.991 ± 0.12
Simple Pendulum/OOP/BoundaryValueDiffEqMIRK.MIRK6()	1.6 ± 0.29 ms	1.6 ± 0.32 ms	0.999 ± 0.27
Simple Pendulum/OOP/MultipleShooting(10, Tsit5; grid_coarsening = false)	3.4 ± 2.5 ms	3.37 ± 2.6 ms	1.01 ± 1.1
Simple Pendulum/OOP/MultipleShooting(10, Tsit5; grid_coarsening = true)	6.45 ± 4.9 ms	6.46 ± 4.8 ms	1 ± 1.1
Simple Pendulum/OOP/MultipleShooting(100, Tsit5; grid_coarsening = false)	0.0875 ± 0.004 s	0.0925 ± 0.0047 s	0.946 ± 0.065
Simple Pendulum/OOP/MultipleShooting(100, Tsit5; grid_coarsening = true)	0.134 ± 0.0086 s	0.142 ± 0.01 s	0.938 ± 0.091
Simple Pendulum/OOP/Shooting(Tsit5())	0.524 ± 0.056 ms	0.527 ± 0.14 ms	0.995 ± 0.28
time_to_load	6.58 ± 0.086 s	6.82 ± 0.11 s	0.965 ± 0.02

### Benchmark Plots A plot of the benchmark results has been uploaded as an artifact to the workflow run for this PR. Go to "Actions"->"Benchmark a pull request"->[the most recent run]->"Artifacts" (at the bottom).