The latest push should finally get all tests passing again.

Internal time-keeping is now all set up. Every process now has an attribute self.current_time which is kept up to date, and also synchronized down the subprocess tree. It is also possible to initialize a process with an arbitrary initial_time, and/or reset the clock of a process and its subprocesses by assigning a new np.datetime64 object to self.current_time after process creation.

Everything seems to be working as it should now.

The next step is to get back to a more efficient insolation calculation. Currently we are calling the insolation calculator at every timestep. It is much slower than the previous strategy of pre-computing insolation for every timestep throughout a calendar year.

It would be great to have a parent class that handles this kind of logic for any type of diagnostic process that depends only on time, so it can be precomputed year by year.

Also need to write some documentation about the new time and date handling.

The latest push restores performance by pre-computing insolation one year at a time. The logic for this is now implemented in climlab.radiation.DailyInsolation by just matching the current time to a precomputed list of times.

It would be very simple to abstract this logic out to a parent class for diagnostic processes that depend on time alone. But I don't have any particular application for this right now so I'll leave it as is.

Everything is working and the docs are written. This is ready to merge. Major new functionality for climlab v0.10