Add support for configurable ONNX SessionOptions across feature extraction and inference modules

[dalgwen]

Allows, for example, to set up more conservative CPU usage like this:

    import onnxruntime as ort
    opts = ort.SessionOptions()
    opts.intra_op_num_threads = 1
    opts.inter_op_num_threads = 1
    opts.execution_mode = ort.ExecutionMode.ORT_SEQUENTIAL
    opts.add_session_config_entry("session.intra_op.allow_spinning", "0")
    opts.add_session_config_entry("session.inter_op.allow_spinning", "0")

    model = WakeWordModel(models=model_files, sess_options=opts)

These example session options drastically reduce CU consumption by mono-threading everything and using sleep instead of spinning/busy waiting.

Suitable for a 'softer' background process in a small environment. Cost is theoretically higher latency (not humanly noticeable).