Fixup broken Windows CI

[bpkroth]

pytest is failing in Windows atm.

Here's an example: https://github.com/microsoft/MLOS/actions/runs/5352292739/jobs/9707056812

mlos_core/mlos_core/tests/optimizers/bayesian_optimizers_test.py::test_context_not_implemented_error[RandomOptimizer-kwargs0] 

mlos_core/mlos_core/tests/spaces/adapters/llamatune_test.py::test_llamatune_pipeline[8-special_param_values0-50] 

[gw0] PASSED mlos_core/mlos_core/tests/optimizers/bayesian_optimizers_test.py::test_context_not_implemented_error[RandomOptimizer-kwargs0] 

mlos_core/mlos_core/tests/optimizers/bayesian_optimizers_test.py::test_context_not_implemented_error[EmukitOptimizer-kwargs1] 

[gw1] PASSED mlos_core/mlos_core/tests/spaces/adapters/llamatune_test.py::test_llamatune_pipeline[8-special_param_values0-50] 

mlos_core/mlos_core/tests/spaces/adapters/llamatune_test.py::test_llamatune_pipeline[8-special_param_values1-250] 

[gw1] PASSED mlos_core/mlos_core/tests/spaces/adapters/llamatune_test.py::test_llamatune_pipeline[8-special_param_values1-250] 

mlos_core/mlos_core/tests/spaces/adapters/llamatune_test.py::test_llamatune_pipeline[8-special_param_values2-50] 

[gw0] FAILED mlos_core/mlos_core/tests/optimizers/bayesian_optimizers_test.py::test_context_not_implemented_error[EmukitOptimizer-kwargs1] 



================================== FAILURES ===================================

_________ test_context_not_implemented_error[EmukitOptimizer-kwargs1] _________

[gw0] win32 -- Python 3.10.11 C:\Miniconda\envs\mlos\python.exe



configuration_space = Configuration space object:

  Hyperparameters:

    x, Type: UniformFloat, Range: [0.0, 1.0], Default: 0.5

    y, Type: Categorical, Choices: {a, b, c}, Default: a

    z, Type: UniformInteger, Range: [0, 10], Default: 5



optimizer_class = <class 'mlos_core.optimizers.bayesian_optimizers.emukit_optimizer.EmukitOptimizer'>

kwargs = {}



    @pytest.mark.parametrize(('optimizer_class', 'kwargs'), [

        *[(member.value, {}) for member in OptimizerType],

    ])

    def test_context_not_implemented_error(configuration_space: CS.ConfigurationSpace,

                                           optimizer_class: Type[BaseOptimizer], kwargs: Optional[dict]) -> None:

        """

        Make sure we raise exceptions for the functionality that has not been implemented yet.

        """

        if kwargs is None:

            kwargs = {}

>       optimizer = optimizer_class(parameter_space=configuration_space, **kwargs)



configuration_space = Configuration space object:

  Hyperparameters:

    x, Type: UniformFloat, Range: [0.0, 1.0], Default: 0.5

    y, Type: Categorical, Choices: {a, b, c}, Default: a

    z, Type: UniformInteger, Range: [0, 10], Default: 5



kwargs     = {}

optimizer_class = <class 'mlos_core.optimizers.bayesian_optimizers.emukit_optimizer.EmukitOptimizer'>



mlos_core\mlos_core\tests\optimizers\bayesian_optimizers_test.py:30: 

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

mlos_core\mlos_core\optimizers\bayesian_optimizers\emukit_optimizer.py:44: in __init__

    from emukit.examples.gp_bayesian_optimization.single_objective_bayesian_optimization import GPBayesianOptimization

        __class__  = <class 'mlos_core.optimizers.bayesian_optimizers.emukit_optimizer.EmukitOptimizer'>

        parameter_space = Configuration space object:

  Hyperparameters:

    x, Type: UniformFloat, Range: [0.0, 1.0], Default: 0.5

    y, Type: Categorical, Choices: {a, b, c}, Default: a

    z, Type: UniformInteger, Range: [0, 10], Default: 5



        self       = EmukitOptimizer(parameter_space=Configuration space object:

  Hyperparameters:

    x, Type: UniformFloat, Range: [0.0,... 0.5

    y, Type: Categorical, Choices: {a, b, c}, Default: a

    z, Type: UniformInteger, Range: [0, 10], Default: 5

)

        space_adapter = None

C:\Miniconda\envs\mlos\lib\site-packages\emukit\examples\gp_bayesian_optimization\single_objective_bayesian_optimization.py:8: in <module>

    from GPy.kern import Matern52

        Enum       = <enum 'Enum'>

        __builtins__ = <builtins>

        __cached__ = 'C:\\Miniconda\\envs\\mlos\\lib\\site-packages\\emukit\\examples\\gp_bayesian_optimization\\__pycache__\\single_objective_bayesian_optimization.cpython-310.pyc'

        __doc__    = None

        __file__   = 'C:\\Miniconda\\envs\\mlos\\lib\\site-packages\\emukit\\examples\\gp_bayesian_optimization\\single_objective_bayesian_optimization.py'

        __loader__ = <_frozen_importlib_external.SourceFileLoader object at 0x00000222678F6[410](https://github.com/microsoft/MLOS/actions/runs/5352292739/jobs/9707056812#step:11:411)>

        __name__   = 'emukit.examples.gp_bayesian_optimization.single_objective_bayesian_optimization'

        __package__ = 'emukit.examples.gp_bayesian_optimization'

        __spec__   = ModuleSpec(name='emukit.examples.gp_bayesian_optimization.single_objective_bayesian_optimization', loader=<_frozen_imp...envs\\mlos\\lib\\site-packages\\emukit\\examples\\gp_bayesian_optimization\\single_objective_bayesian_optimization.py')

        np         = <module 'numpy' from 'C:\\Miniconda\\envs\\mlos\\lib\\site-packages\\numpy\\__init__.py'>

[bpkroth]

#423