Small bugfix for the TimeSeasonality component.

In the classical definition, the n-th seasonal effect isn't identified, so n-1 effects are defined and the transition matrix is:

$$ T= \begin{bmatrix} -1 & -1 & -1 & -1 \\ 1 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 1 & 0 \\ \end{bmatrix} $$

It's no obvious, so there's a worked example in the API docs for why this does what you want it to do.

I wanted to add an argument that would allow users to handle this normalization themselves if they wished, but it was implemented incorrectly. Now it should work. If you pinky swear that the seasonal effects are normalized (e.g. with ZeroSumNormal prior), you can get a transition matrix like:

$$ T = \begin{bmatrix} 0 & 1 & 0 & 0 \\ 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & 1 \\ 1 & 0 & 0 & 0 \end{bmatrix} $$

Which will just loop the effects around without doing anything fancy. This will be much easier to interpret.