Allow full tensorflow if tflite_runtime isn't available

[mikejgray]

I have a bit of a strange use case, where I'm trying to run OpenWakeWord on Apple Silicon and tflite-runtime isn't available for it. I am running tensorflow with tensorflow-metal to take advantage of the chip's capabilities, but OWW doesn't attempt to fall back to tensorflow, instead opting for onnx.

It would be nice for full tensorflow to be an option, even if it's just a configurable option with extra dependencies.

Thank you!

[mikejgray]

I haven't tested extensively, but it looks like an additional except ImportError in model.py and utils.py does the trick:

from tensorflow.lite.python import interpreter as tflite

[dscripka]

That's a good solution I think, as managing dependencies across many different platforms can be quite complex. I'm hesitant to add tensorflow as a dependency for the base package as it's so large, but some logic that would attempt imports if packages are available (perhaps using find_spec) could work. I'll consider this for the next release.

[mikejgray]

Definitely no need to add tensorflow as a dependency, but supporting environments that use it would be fantastic. :)