Average predictions from the 4 ML models to find start and end times of orca calls

[Benjamintdk]

Resolves #42

• Refactored predict.py by creating a separate function, generate_predictions, to account for cases where MODEL_PATH is either a file or directory path to saved weights, and averages the results obtained for the latter
• Added functionality at the end of the predict function to save the labels of high and low confidence examples to a csv file
• Updated ReadMe to explain changes made