Dopair1 vectorisation merge

Summary
-------
- Contains a vectorised version of `runner_dopair1_density`
- `runners` now have two particle caches to hold cell ci and cell cj
- The `cache` has been updated to include `max_d` which is the maximum distance into the neighbouring cell
- `cache.h` contains more functions to read two cells into cache and to read a subset of particles from each cell into the cache


See merge request !320

Added softened gravity to the M2L kernels



See merge request !336

Crucial +1 for the edge case where all particles in cell ci interact with at least one particle in cell cj.

Updated runner_doself1_density_vec and runner_dopair1_density_vec to align with Pedro's recent changes in DOPAIR1 and DOSELF1.

Do not clear the sort flag when receiving parts after the ghost.

Implements #304. We only reset the c->sorted flag for the first receive not for the second one. 

See merge request !335

Do not clear the sort flag when receiving parts after the ghost. We can re-use the sort from the density.

Removed the 'tight' property of the tasks. Only density pairs need to be tight so we can just use that.

Add a separate timer for subset sub-cell tasks. Write the timers to a file. Make sure we can compile without timers. Add a configure option to switch the timers on.

Merge branch 'dopair-vectorisation-merge' of gitlab.cosma.dur.ac.uk:swift/swiftsim into dopair-vectorisation-merge

Conflicts:
	src/runner_doiact.h

Rebuild criteria

Instead of rebuilding whenever the cell constraints or sort indices are violated, just re-sort in case of the latter.

This lead to some other changes:

 * Init tasks are gone, particle initialization is now handled in the drift tasks,
 * Drift tasks no longer act on the super-cell, but exist at every level similar to the sort tasks,
 * Fixed several dependency issues with sort and drift tasks,
 * Only drift/sort on-demand in the sub-cell tasks.

Marked as [WIP] as this has not yet been tested for MPI runs, or with gravity.

See merge request !327