Support Apple Silicon via CoreML

[joefutrelle]

This pull request enhances device selection logic for inference and adds comprehensive tests for provider selection. The main improvements are support for Apple's CoreML as a provider, improved handling of the --cpuonly flag, and a new test suite to verify provider selection behavior.

Device/provider selection improvements:

• Updated the --cpuonly argument help text to clarify that it disables both CUDA and CoreML providers, not just CUDA.
• Improved the get_providers function in cli.py to:
  • Prefer CUDA if available, otherwise select CoreML with configurable options, and finally fall back to CPU.
  • Respect the --cpuonly flag to always force CPU, regardless of other providers.

Test coverage:

• Added TestGetProviders test class in tests/test_infer_functions.py to cover all combinations of provider availability and environment overrides, ensuring correct provider selection logic.
• Imported get_providers into the test file for direct testing.

Please note that there is E5RT-related logspam that this PR doesn't suppress--the issue it identifies does not affect performance or output.