ModelingToolkitBase: allow DomainSets 0.8

[ChrisRackauckas-Claude]

Please ignore until reviewed by @ChrisRackauckas.

What

Widen the ModelingToolkitBase DomainSets compat bound from 0.6, 0.7 to 0.6, 0.7, 0.8 (and bump the version 1.42.1 -> 1.42.2).

Why

This is required to unblock the DomainSets 0.7 -> 0.8 dependabot bump in MethodOfLines.jl (SciML/MethodOfLines.jl#564). While diagnosing that PR I found that PDEBase was not the only blocker: ModelingToolkitBase also pins DomainSets = "0.6, 0.7", so any package that pulls ModelingToolkit (PDEBase, MethodOfLines, ...) cannot resolve DomainSets 0.8. Forcing DomainSets 0.8 in a MethodOfLines test environment produces:

ERROR: Unsatisfiable requirements detected for package ModelingToolkitBase [7771a370]:
 └─restricted by compatibility requirements with DomainSets [5b8099bc] to versions: uninstalled
   └─DomainSets [5b8099bc] restricted to versions 0.8 by an explicit requirement, leaving only versions: 0.8.0

Symbolics already supports DomainSets 0.8 (DomainSets = "0.6 - 0.8" from Symbolics 7.18.1), so ModelingToolkitBase is the last piece of the core MTK stack still capping it at 0.7.

DomainSets 0.8 breaking changes and why they don't affect ModelingToolkitBase

DomainSets v0.8.0 (JuliaApproximation/DomainSets.jl#178):

1. Removed an old type piracy so DomainSets.× and LinearAlgebra.× are now distinct functions.
2. Unexported some internal symbols that now live in FunctionMaps.jl.

ModelingToolkitBase's only DomainSets usage is in lib/ModelingToolkitBase/src/utils.jl (and import DomainSets in the module):

const AnyInterval{T} = DomainSets.Interval{A, B, T} where {A, B}
...
lo, hi = unwrap.(DomainSets.endpoints(domain))
lo, hi = DomainSets.endpoints(domain)

DomainSets.Interval and DomainSets.endpoints are fully qualified and remain exported/defined in DomainSets 0.8 (verified locally below). No bare × and no unexported internals are used, so no source changes are required — only the compat bound.

Local verification

DomainSets 0.8.0 provides every name ModelingToolkitBase uses:

DomainSets version: 0.8.0
endpoints defined: true
endpoints(iv) = (0.0, 2.0)
Interval ctor works: 0.0 .. 2.0
AnyInterval alias ok; iv isa AnyInterval: true

With this ModelingToolkitBase branch + a DomainSets-0.8-compatible PDEBase developed into a MethodOfLines test environment, resolution succeeds and selects DomainSets 0.8.0, ModelingToolkitBase 1.42.2, and the full MTK/OrdinaryDiffEq/NonlinearSolve stack precompiles cleanly. ModelingToolkitBase precompiled successfully against DomainSets 0.8.0 (the pre-existing @nospecialize annotation only supported on the first 32 arguments warning is unrelated to this change). See the MethodOfLines PR for the full passing test output.

🤖 Generated with Claude Code

---

Part of the DomainSets 0.8 set: SciML/MethodOfLines.jl#565 (consumer), SciML/PDEBase.jl#81 (the other 0.7 pin).