Raise OrdinaryDiffEq solver extras floors to 2

[ChrisRackauckas-Claude]

Please ignore this PR until it has been reviewed by @ChrisRackauckas.

What changed and why

Pure [compat] lower-bound bumps in extras:

• OrdinaryDiffEqBDF = "1, 2" → "2"
• OrdinaryDiffEqDefault = "1, 2" → "2"
• OrdinaryDiffEqRosenbrock = "1, 2" → "2"
• OrdinaryDiffEqSDIRK = "1, 2" → "2"

OrdinaryDiffEqNonlinearSolve = "1.15, 2" is unchanged (already floored).

No code, tests, or package version changes.

The package requires ModelingToolkit = "11.28", which only allows SciMLBase 3. The 1.x line of these solver packages only allows SciMLBase 2, so the declared "1, 2" extras floor is dead. Rosenbrock 1.0.0 is also not a registered version (1.x starts at 1.1.0).

Verification (Julia 1.10.11)

Pkg.develop of ModelingToolkitStandardLibrary, then Pkg.add(name=..., version=...):

# Rosenbrock 1.0.0 (old 1.x floor)
RESOLVE_FAIL OrdinaryDiffEqRosenbrock 1.0.0
Unsatisfiable requirements detected for package OrdinaryDiffEqRosenbrock
 ├─possible versions are: 1.1.0-2.6.5 or uninstalled
 └─restricted to versions 1.0.0 by an explicit requirement — no versions left

# BDF 1.0.0 / Default 1.0.0
RESOLVE_FAIL (same: 1.x starts at 1.1.0)

# Rosenbrock 2.0.0 (new floor)
RESOLVE_OK OrdinaryDiffEqRosenbrock 2.0.0

What was not verified

• Full MTKSL test suite under the pinned 2.0.0 graph (only develop + add).
• GPU.
• 1.1.0 of each solver as an intermediate pin (1.x is SciMLBase 2-only against ModelingToolkit 11.28).

Links

• Same solver 1.x / SciMLBase 3 split: Raise OrdinaryDiffEqTsit5 floor to 2 and ModelingToolkit to 11.21 Corleone.jl#123