Update MATLAB requirement from 0.8 to include 0.9

[ChrisRackauckas-Claude]

Summary

• Updates the MATLAB.jl compat bound from 0.8 to 0.8, 0.9 to allow using MATLAB.jl v0.9.0

Changes

The only change is to Project.toml to add MATLAB.jl 0.9 as a compatible version.

Breaking Changes in MATLAB.jl 0.9

According to the MATLAB.jl v0.9.0 release notes:

1. AbstractArray Conversion: Types that are subtypes of AbstractArray (e.g. Diagonal) are now converted to MATLAB arrays instead of structs.
2. Tuple Handling: Tuples are no longer converted to MATLAB dictionaries. Instead tuples are converted to cell arrays (if it has mixed types) or numeric arrays (if all elements share a single numeric type).

These breaking changes do not affect MATLABDiffEq.jl because:

• The package passes regular arrays (vectors) for parameters, initial conditions, and time spans
• All the APIs used by MATLABDiffEq.jl (put_variable, get_variable, eval_string, jvector, jarray, get_mvariable, get_default_msession) remain unchanged

Key Improvements in MATLAB.jl 0.9

• The segfault issue with MATLAB versions 2022/2023 is gone on Linux platforms
• Improved Windows and Linux path handling
• Expanded AbstractArray support

Fixes #57

cc @ChrisRackauckas

🤖 Generated with Claude Code

[ChrisRackauckas-Claude]

The CI test failure is expected since MATLAB is not available on the CI runner. This is the same behavior as previous PRs (#54, #56) which were merged despite the test failures.

The only change in this PR is updating the MATLAB compat bound from 0.8 to 0.8, 0.9. I verified that:

1. The package resolution succeeds with MATLAB.jl v0.9.0
2. All APIs used by MATLABDiffEq.jl are still exported in MATLAB.jl 0.9
3. The breaking changes in MATLAB.jl 0.9 (AbstractArray conversion and tuple handling) don't affect this package since it only passes regular arrays

cc @ChrisRackauckas