Add more complex examples to tutorials showcasing differentiability

[gomezzz]

Feature

Desired Behavior / Functionality

torchquad allows fully differentiable numerical integration. This can enable neural network training through integrals. This capability deserves a dedicated example. There is a example on the gradient computations in the docs already. However, training a simple neural network on e.g. a gravitational potential or similar problem could be a cool example?

What Needs to Be Done

• Find a nice example to showcase neural network training / optimization capabilities due to torchquad's differentiability
• Add to docs