Don't remove rshift from hi_max and pass h_max to populate_max_index_no_cache_force. Fixes missing interactions in runner_dopair2_force_vec.

Add a print units

See merge request !463

Remove unnecessary dependency between sort and self/density tasks.

See merge request !461

Improvements to i/o and parallel-i/o

See merge request !460

Corrected typo in Minimal-SPH implementation that prevents neighbour number convergence in some cases.

Make sure the code compiles with vectorization switched on and using a different hydro scheme from Gadget-2.

Real problem was a 28 * k, instead of k < 26 *, but clear up the logic and fix some edge cases where the range could over/underflow a bit, also add a debugging check to spot this issue next time

Add time based edge repartitioning

See merge request !458

Task graph

Closes #92

See merge request !459

Use non-buffered MPI sends for small messages

See merge request !457

Implement more general conversion functions for i/o. Can now add one that does box-wrapping  of particles.

Conflicts:
	examples/process_cells
	examples/process_cells_helper
	src/debug.c
	src/debug.h
	src/partition.c

Sort out naming of output cell dumps to be based on the step, so they can be compared to other dumps