Fix issue with attempt to dump partition when not using MPI

Can also change the number of queues to match the threads (expected behaviour), the number of tasks per cell and the size of MPI messages, these all could be useful when re-starting

Exits the application and writes restart files (regardless of other settings)

Sugar of various kinds, new parameters to control restart file dumping, better adherence to the restart_ name space, dumps on exit for testing restarts, include in step properties

Remove debugging output, tidy up output and make sure we return restart files in order, at least up to 1M files.

Needed to save and restore the current partition

Moved restart_write into engine_step and trigger after drift_all

Tidies up the restart option

Start reorganizing so we can restart, doesn't work yet

Add the possibility to read basic chemistry information from the parameter file and propagate this onto the particle at initialisation time. More uniform naming convention for the first initialisation of properties.

Grackle chemistry --> GEAR chemistry. Moved the output of subgrid model to a separate part of the HDF5 file.

This reverts commit d6acb7a1.

Meant to add to another branch...

Make use of a temporary threadpool in the reading routines to speed-up the duplication of parts into gparts.

Open only once dependency file and add an option in debugging-checks mode for writing task dependency graph

Add information to stdout and timestep.txt about what happened in the step (rebuild, repartition, snapshot, ...)

When compiling with the iosthermal equation of state, read the constant thermal energy from the YAML file.

There is a problem when the cpu mask is empty and using multiple MPI ranks on a node, in that case all the main threads end up pinned to one core during the reading of the initial condition (and any other tasks done before engine_init is called).