Can use json to pass state variable tolerances.
Can also just pass in atol and rtol to have all state variables set the same.

Moved the calls to fortran files to a singular file so we don't have to go
through C++ and Fortran for multiple functions.

This is now rectified

Moved read_attrb and read_icond to be called from job_init_fotran

 summa_SetTimesDirsAndFiles
 summa_defineGlobalData

The actors version now calls job_init_fortran as an interface which calls the SUMMA supplied subroutines.

Replaced some messages with request statements.
This is to show the direct relationship the actors have in this request, instead
of relying on a send and send.

Used requests where it made sense in the actors. This is where the HRU
would send a message then wait for a response anyway. Instead of defining
messages for these cases we use request.

mDecisions would return an error, however the C++ side had no mechanism to
view or inspect the contents of the message.

A interface function in fortran was created to print the message.

HRU can now restart with timestep halved.

Changed the GRUinfo class to be just the GRU class.
This class now better reflects that it is the object that
the job actor holds. It will also contain the GRU actor that runs the simulation
portion of the program.

HRU timing was slowing down the code. It has now been removed.
The wall-clock-time of each hru is now the metric recorded.

Removed the reading of json variables that were writing to csv.

Added a debug mode for kinsol.

When an actor fails it is now is noted and writen to netCDF that it
failed. Restarting the actor will be a little
tricky but this should allow for the running of Summa-Actors to see how
many HRUs fail.

Added functionality for the job actor to record the stats of GRU actors
When the simulation is finished these stats are passed to the
file access actor where they are written to netcdf

Can disconnect client anytime during the simulation. Can also disconnect the backup servers. Can add a backup server

 - When the server goes down and the backup takes over a client going down would not be noticed by the backup server.
 - The problem was the backup server was never monitoring the clients. This is now fixed

Can get 1 HRU managed by the file access actor. An HRU will wait until it is told it can write again.