ColBert is not failing when Error is encounter during both train and indexing

[AnastasiiaShukhova]

When dealing with Colbert I ran many times into an issue when the code breaks, the error message is logged and not propagated, but the cell in Jupyter notebook still executes. It seems like the error handling has been broken.

When running the code on Google Colab log messages have been supressed and I couldnt see any error messages at all, only lonely "Starting...".