Transcription Model Configuration per Device

[bennettrwu]

The performance parameters of transcription models should be configured on a per-device basis rather than in code via model_factory.

• Each model should be modified to take in a model_configuration object rather than arguments.
  • model_configuration has no required schema other than that it can be configured via device_configuration (later).
  • model_implementations should use this to configure all configurable settings. (Please document schema)
  • As a result, every model_implementation should only require 2 parameters (websocketandmodel_configuration`)
• Create a new device_configuration config file that contains mappings from model_key -> (model_implementation, model_configuration)
• model_factory.py should be modified to lookup model_key in device_configuration rather than in match statement. model_factory.py can then create an instance of model_implementation using model_configuration.