Add StrangMarchuk second-order symmetric operator splitting

[oameye]

Summary

Implements the classical Strang-Marchuk ABA splitting scheme as requested in #10. For two operators A and B, the scheme performs A(Δt/2) → B(Δt) → A(Δt/2), achieving second-order accuracy through symmetry.

• Added StrangMarchuk algorithm struct with inner constructor validating exactly 2 operators
• Cache, init_cache, and _perform_step! in src/solver.jl
• Exported StrangMarchuk from the main module
• Handles floating-point time drift from two half-steps via try_snap_children_to_tstop!
• Handles instability detection (NaN propagation between sub-steps) gracefully
• Added Strang 1968 and Marchuk 1971 bibliography entries
• Added StrangMarchuk precompilation workload
• Updated docs: API reference, usage example, and theory section with convergence analysis

References

• G. Strang, On the construction and comparison of difference schemes, SIAM J. Numer. Anal. 5 (1968), pp. 506–517
• G. I. Marchuk, On the theory of the splitting-up method, in Numerical Solution of Partial Differential Equations-II, Academic Press, 1971, pp. 469–500

Closes #10

Test plan

• All existing LieTrotterGodunov tests pass unchanged (1736 total)
• StrangMarchuk integrated into existing test loops alongside LTG
• Convergence order test with non-commuting operators confirms first-order (LTG) vs second-order (SM)
• Nested splitting tests (StrangMarchuk wrapping StrangMarchuk)
• Adaptive solver wrapper tests (FakeAdaptiveSM)
• Reinit and TimeChoiceIterator tests
• Instability detection tests (NaN propagation)
• Full test suite passes locally

[termi-official]

Formatting is needed and I left a comment.

[termi-official]

LGTM. Thanks for ironing out the details here and adding the order tests!!

[oameye]

Can we tag a release?

[termi-official]

@oscardssmith