Make mkl-service an optional dependency

[ndgrigorian]

mkl-service is only a dependency for the sake of the scipy interface, which is optional

This PR proposes moving mkl-service to optionally installable and adding logic to check for both mkl-service and scipy in the environment before exposing the scipy interface

[Copilot]

Pull request overview

This PR makes mkl-service optional by moving it into the scipy_interface/test extras and conditionally exposing the SciPy adapter only when both SciPy and the mkl Python module (from mkl-service) are present.

Changes:

• Moves mkl-service out of core dependencies and into optional extras in pyproject.toml.
• Updates mkl_fft.interfaces to only import/export scipy_fft when both SciPy and mkl-service are available.
• Adjusts conda recipes and CI workflows to install mkl-service explicitly for SciPy-related testing.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
pyproject.toml	Shifts mkl-service into extras and changes base dependency set.
mkl_fft/interfaces/__init__.py	Adds runtime checks to gate the SciPy interface export.
conda-recipe/meta.yaml	Moves runtime MKL dependency to mkl and adds mkl-service to test requirements.
conda-recipe-cf/meta.yaml	Same dependency adjustments for conda-forge recipe.
.github/workflows/conda-package.yml	Ensures mkl-service is installed for test envs that include SciPy.
.github/workflows/conda-package-cf.yml	Same test-environment adjustments for conda-forge workflow.
.github/workflows/build_pip.yaml	Updates MKL install and adds explicit mkl-service install for tests (currently redundant with extras).

[Copilot]

find_spec("scipy") only checks that a module is discoverable, not that it can actually be imported. If SciPy is present but import fails at runtime (e.g., missing shared libs / ABI issues), this will still attempt to import scipy_fft and can make import mkl_fft.interfaces fail. Consider reverting to a try: import scipy (and similarly import mkl) guard like the previous implementation so failures are safely ignored.