Make confidence threshold configurable in nn_train

[Saagar-Arya]

Summary

The confidence threshold is currently a hard-coded value in nn_train. To improve flexibility and usability, this threshold should be made configurable via a parameter, environment variable, or configuration file.

Motivation

Allowing users to set the confidence threshold enables better control over model training and evaluation, aligning with different dataset requirements or experiment goals.

Proposal

• Refactor the nn_train component so the confidence threshold can be set externally by users.
• Documentation should be updated to describe how to set the new configurable parameter.

Acceptance Criteria

• The confidence threshold is no longer hard-coded and can be set by the user.
• Unit tests and documentation are updated to reflect this change.