add_activations_recorder

[Sentz98]

This PR introduces the ability to record activations from specific model layers during testing.

Activations are saved in a CSV file for further analysis.
Headers include: Layer Name, Layer Type, Activation Shape, and Activations.

Forward hooks are automatically registered on all leaf modules during testing.
Hooks are automatically removed when recording ends.

Usage Example:

model.enable_activation_recording("path/to/output.csv")
# Run testing loop better if only with one image
model.close_activation_file()

[fpaissan]

Thanks for the contribution. While I can see the value in having such API implemented, I'm not sure of the technical details:

• saving all activations can result in a huge memory usage, depending on the network activations
• why do you think csv is the best option for the task?