CPU time fraction. Also allow the balance test to be done if fewer
than the complete number of particles interact. Finally capture the
CPU times to a file so that the balance can be assessed. This is
only done when task timings are also enabled.

Conflicts:
	examples/main.c
	src/engine.c
	src/engine.h

Compiles, but not tested.

Timestep order

This is my last batch of changes regarding time integration.
 
 - Additional checks here and there that the particles are synchronized. Especially gpart.
 - Re-arranged the time-step sequence and simplified the logic in engine_step().
 - Moved the decision of repartitioning from the main() to engine_step().

The re-arrangement of the time-step sequence allows for perfect energy conservation even when we rebuild, repartition or dump snapshots. Previously that was not the case as a spurious drift had to be introduced. 

See merge request !314

Properties of the self-gravity scheme are now read from the YAML file. They are printed to stdout and to snapshots.

Properties of the self-gravity scheme are now read from the YAML file. They are printed to stdout and to snapshots.

Added the option to replicate the particles along the axis to create artificially larger test cases. Driven by the parameter 'InitialConditions:replicate'.

No unnecessary loop in pair

This implements two tiny improvements:
 
 - In the PAIR task, only loop over the particles in a cell that has active particles.
 - In the ghost task, construct the list of indices of active particles only. No need to loop over the inactive ones.

See merge request !313