Allow getting derivative at arbitrary time points in tendency

[SBlokhuizen]

This PR rewrites how derivatives are handled in the tendency classes. Before, you could only get the derivatives at the start or end of the tendency. This implementation extends this so the derivative can be obtain at any point in the tendency time range.

This might be necessary for example when using a repeated tendency, followed by a smooth tendency. The repeated tendency can end at any point in a tendency, and the smooth tendency should be able to match the derivative for the last tendency. For example:

Main changes in this PR:

• Introduction of of get_derivative/get_value methods, these replace the get_start_value, get_end_value, get_start_derivative and get_end_derivative methods.
• There were some issues due to the ambiguity of passing/returning floats, lists, and numpy arrays to the get_value/get_derivative methods. For clarity, I have changed these functions to only take in numpy arrays, and only output numpy arrays, at the expense that you have to map these back to a single float if only the at a single point is required.

[maarten-ic]

Nice start! I've got a couple of comments to polish the PR, let me know if anything's unclear 🙂

[maarten-ic]

Let's move this to a fixture that's automatically used by all tests in this file?

@pytest.fixture(autouse=True)
def patch_periodic_base_tendency():
    arr = np.array([0])
    with (
        patch.object(PeriodicBaseTendency, "get_value", return_value=(arr, arr)),
        patch.object(PeriodicBaseTendency, "get_derivative", return_value=arr),
    ):
        yield