Typing of scalar derivatives of state variables

[mlprt]

feedbax.dynamics.AbstractDynamicalSystem is a generic of StateT. Each subclass must define a vector field that returns the time derivatives of each array in a state PyTree.

def vector_field(
    self, 
    t: float, 
    state: StateT, 
    input: PyTree[Array],  
) -> StateT:
    """Returns the time derivatives of the system's states."""
    ...

The return type StateT is more or less correct: the return value will have the same PyTree structure as state, and -- because the time derivative is scalar -- the same array shapes as well.

This seems fine to me, but perhaps there is a better way to make it explicit when we are passing around derivatives.

[mlprt]

This will probably be obviated if we replace the staged model approach with a DAG approach (#28), so I'm tagging this issue as "invalid".