[BUG] Wrong channel number for 5G in neural network training

When using a 5G device, the neural network does not take into account the number of channels. Basically in epoch.py and reward.py, everytime that it uses `wifi.NumChannels` it should be using `wifi.NumChannelsExt` on 5G devices. Probably needs a similar solution to what was done in featurizer.py

Start by comparing the existing handling in featurizer.py with the uses of wifi.NumChannels in epoch.py and reward.py. Verify which channel value is used for 5G devices, then check that the affected training and reward paths consistently use the extended value where appropriate.

Written by the indexing model from the issue text.