[Docs] Documentation improvements

[jpmoutinho]

Here we list a few things that need to be made more explicit in the docs:

• The usage of set_trainable to change between feature and variational parameters in blocks. Maybe can be added in the Parameters section.
• A few things on the usage of QuantumModel:
  • How to reset quantum model parameter values to specific value (model.reset_vparams(...))
  • How to get derivatives with respect to trainable parameters (e.g., autograd.grad(loss, qnn.parameters())).
  • How to obtain the parameter dict at a given training epoch (model._params)