Fix discretized-DAE tests/docs by passing initializealg explicitly per-call (no library default change)

[ChrisRackauckas-Claude]

Please ignore until reviewed by @ChrisRackauckas.

What this does

Fixes the master-CI reds (DAE, Higher_Order, MOL_Interface1 test groups, plus the Schrödinger docs example) without changing MethodOfLines' default DAE initialization. MethodOfLines keeps its existing default (downstream CheckInit). Each failing case instead passes initializealg = BrownFullBasicInit() explicitly in its own solve(...) call.

These are MOL discretizations of PDEs with algebraic constraints (elliptic coupling, over-determined boundaries, Robin/flux BCs, complex split into real/imag). They are built with build_initializeprob = false, so the solver's default CheckInit rejects the finite-difference-inconsistent ICs with:

DAE initialization failed: your u0 did not satisfy the initialization requirements ... please pass initializealg = BrownFullBasicInit() to solve

BrownFullBasicInit holds the differential variables fixed and solves only the algebraic equations.

Reverted (vs the earlier version of this PR)

• src/interface/MOLFiniteDifference.jl: the block injecting initializealg = BrownFullBasicInit() into the discretization kwargs is fully reverted — this file is now identical to master. No library default change.
• src/MOL_discretization.jl: the _problem_opts... positional absorber and the func_kwargs = Base.structdiff(...) init-strip (which only existed to support the kwargs injection) are reverted; the ODEFunction(pdesys, discretization; ...) signature and discretization.kwargs... splat are restored.

Independent source bugfix kept (justified)

• src/MOL_discretization.jl: f_analytic = nothing is initialized before the if analytic !== nothing block. On master, ODEFunction(simpsys; analytic = f_analytic, ...) references f_analytic unconditionally while it is only assigned inside the if, so ODEFunction(pdesys, disc) with no analytic function is an UndefVarError. This is a real pre-existing bug independent of any init change, so the one-line guard is kept.

Per-call explicit fixes

File	failing case	fix
test/DAE/MOL_1D_PDAE.jl	elliptic 0 ~ Dxx(v) + ... constraint	solve(prob, Rodas4(), saveat = 0.1, initializealg = BrownFullBasicInit())
test/Higher_Order/MOL_1D_HigherOrder.jl	KdV, over-determined boundary	solve(prob, FBDF(), saveat = 0.1, initializealg = BrownFullBasicInit())
test/MOL_Interface1/MOLtest1.jl	"Rearranged Robin" algebraic Robin BC	solve(prob, Rodas4(), initializealg = BrownFullBasicInit())
docs/src/tutorials/schroedinger.md	complex eq split into real/imag	solve(prob, FBDF(), saveat = 0.01, initializealg = BrownFullBasicInit())

BrownFullBasicInit is imported from DiffEqBase (matching test/Complex/schroedinger.jl, which already does this on master).

Independent docs fixes (not default-init related)

• docs/src/tutorials/PIDE.md: QNDF() → FBDF(). QNDF no longer exists in OrdinaryDiffEq 7 (the docs env resolves OrdinaryDiffEq 7.1.0, where the only exported BDF integrator is FBDF); the block was an UndefVarError. Verified the integro-PDE solves to Success with FBDF and needs no explicit init (its constraint is consistent with the FD ICs).
• docs/src/tutorials/sispde.md (+ docs/Project.toml): add using SteadyStateDiffEq. DynamicSS is no longer exported by DifferentialEquations 8; it lives in SteadyStateDiffEq. This is an API migration, not a default-init change. (The time-dependent Tsit5 solve is pure-parabolic and needs no init.)

Local verification (Julia 1.10; OrdinaryDiffEq 7.1.0, DiffEqBase 7.5.5, MTK 11.28.2, SciMLBase 3.21.0)

Before the per-call fixes (with the src reverted), all three groups error on CheckInit:

• DAE: normresid = 0.00299 > abstol = 1.0e-6
• Higher_Order (KdV): normresid = 0.000158 > abstol = 1.0e-6
• MOL_Interface1 (Rearranged Robin): normresid = 0.559 > abstol = 1.0e-6
• schroedinger docs: normresid = 2182.3 > abstol = 1.0e-6

After the per-call fixes:

• DAE: 22/22 Pass.
• Higher_Order: KdV @test successful_retcode Pass (pre-existing rtol @test_broken unchanged); Beam Test 00 Broken (pre-existing); Test 01 pass.
• MOL_Interface1: every testset green — Heat 1D/2D, Spherical Diffusion, RHS=0, Wave Equation, Rearranged Robin, Array u, Non-Uniform Grid, 2D-1D boundary 2D variable connected to 1D variable via boundary #33.
• docs: schroedinger solves to Success with the explicit init; PIDE and heat Dirichlet/Neumann/Robin solve to Success; sispde uses the SteadyStateDiffEq DynamicSS API.

No tests/examples were skipped, @test_broken-ed, try/catch-swallowed, or tolerance-loosened.

🤖 Generated with Claude Code