Make the VR interface recover the mass of the first high-resolution DM particle to calculate the mean inter-particle separation.

SWIFT can not read DM background particles from the ICs using the particle type 2. These are treated as regular DM gparts after that.

So all stf outputs will increase by 1 and be chronological (0,1,2,3,...n)

The snapshots will also be chronological but will increase by arbitary amounts,
they will share the number of the stf output time they relate to.

So the stfs could go (0,1,2,3,4,5,6,7,8,9) giving us 10 group cataloges. But
there are only 3 snapshots, which could be at any of these number (1,7,8) for
example.

The behaviour when not invoking stf is unchanged. The snapshots and stf outputs
follow their own counts.

A better catch all solution is to track if stf has been run this timestep,
e->stf_this_timestep, and only invoke velociraptor is this value is 0.

outputs at those times (no first and last dumps).

Special case of when you have an output at time_max is dealt with by flagging
that you need an output in the output_list struct, which is then checked at the
end on the last timestep (in main.c).

Also added a check where you cannot run with Snapshots: invoke_stf and
StructureFinder: output_list at the same time (so we dont hit scenarios where
we could run stf twice).

at the snapshot output times), do not use the snapshot basename and instead use
the stf basename (StructureFinding: basename).

This saves the group files to their own folder as is the behaviour when
performing VELOCIraptor using StructureFinding: output_list.

Use the threadpool to populate the array of VELOCIraptor particles from the gparts in the interface.

Pass the correct units to VR. Pass more cosmology info to VR. Only update the particle fields if running VR at a snapshot dump time.

Pass the correct units to VR. Pass more cosmology info to VR. Only update the particle fields if running VR at a snapshot dump time.