Codex/update prediction api error handling and logging

[teehari79]

No description provided.

[roelofsc]

@teehari79 Thank you for the merge request.

Could you please add a summary with a list of changes?
Can you rebase your branch onto the develop branch?

This PR covers a lot of areas, both refactoring and new features. Would it be possible to split up the changes into:

• Bug fixes / improvements / refactoring
• REST API prediction endpoint
• LLM powered narrative endpoint
• Changes to the QuickFaultDetector feature
• Any other added features?

[roelofsc]

@teehari79 The quick_fault_detector does not seem to work with this version anymore. I tested it with the event 44 dataset from the CAREToCompare dataset (quick_fault_detector --c2c_example 'path_to_datasets\Wind_Farm_C\datasets\44.csv').
ARCANA losses are set to 'None' for some detected events (there is an error in 'visualisation.plot_arcana_losses', which is called in 'quick_fault_detection.output.generate_output_plots'). There is no traceback; the last part of the logs is as follows:

configured_ignore_patterns: ('power_2_avg', 'power_5_avg', 'power_6_avg', 'power_17_avg', 'wind_speed_236_avg', 'wind_speed_235_avg', 'wind_speed_237_avg')
1643/1643 [==============================] - 5s 3ms/step
2025-10-20 08:53:30,177 - energy_fault_detector - INFO - Ignoring 7 feature(s) for anomaly scoring (prediction): power_17_avg, power_2_avg, power_5_avg, power_6_avg, wind_speed_235_avg, wind_speed_236_avg, wind_speed_237_avg
2025-10-20 08:53:45,712 - energy_fault_detector - ERROR - An error occurred: 'NoneType' object has no attribute 'columns'

Since the refactoring and bug fixes look promising, I will take a closer look at the individual commits. This will take some time, though.