Star formation (non-MPI)

A bunch of new physics and infrastructure added to the code:

• Remove the sourceterm files. Feedback will be handled differently.
• Recycled the runtime flag -F for star formation.
• When rebuilding, we allocate extra space at the end of the array of each top-level cell for spare particles to be used in on-the-fly creation.
• These particles do not enter any calculation.
• When we want to create a star, we shift all the spart by one slot in order to eat up one of these spares. We update the pointers and counters in that cell hierarchy accordingly.
• That creates a free slot at the start of the cell's array. This is the new particle.
• Added functions to convert a part into a spart.

This should not break anything as we do not use the -F flag by default for now. I can run the EAGLE examples smoothly with the new flag. These tests will become more meaningful when the star formation law is more physically motivated.

The MPI-version will follow soon.

doc/RTD/source/CommandLineOptions/index.rst

@@ -22,7 +22,7 @@ can be found by typing ``./swift -h``::
     -D, --drift-all                   Always drift all particles even the ones
                                       far from active particles. This emulates
                                       Gadget-[23] and GIZMO's default behaviours.
-    -F, --sourceterms
+    -F, --star-formation	      Run with star formation
     -g, --external-gravity            Run with an external gravitational potential.
     -G, --self-gravity                Run with self-gravity.
     -M, --multipole-reconstruction    Reconstruct the multipoles every time-step.