New cosmo ICs

New ICs to replace the BigCosmoVolume at some point.
They contain corrected smoothing lengths such that the initialisation step is much shorter. They also contain the DM particles and further species for when we are ready to have them in SWIFT.

The EAGLE_25 case is almost the old BigCosmoVolume.

See merge request !188

Convert negative weights error in partitioning to warning about tic before toc

A small synchronization error is seen on cosma-e that can be fixed using processory affinity, this stops that being critical.

Fixes #182.

See merge request !192

A small synchronization error is seen on cosma-e that can be fixed using processory affinity, this stops that being critical.

This relives some of the issues in the mpi_skip branch, forgot to push this fix, really part of that

Mpi skip

Fixes #171.

Adds new tasks to propagate cell end times to other nodes so 
that the related tasks can be skipped, avoiding unnecessary
MPI requests.

See merge request !191

Conflicts:
        src/engine.c

Autotools update

Updates for issue #181. Still need to do some testing if tcmalloc is better.

See merge request !190

Useful when profiling.

Previously shared libraries were linked using the unwrapper compiler (so MPI special flags didn't work)

The entropy_in_IC flag is now set and read correctly. Restart from snapshot is possible.

Fix to #178. 

This is a rather temporary fix but allows full compatibility with Gadget. We can now restart from a snapshot and also run for a while with Gadget, then with SWIFT, then with Gadget and so on.

See merge request !189

One major change to replace all gpart id/parts with the new id_or_neg_offset.

As discussed in #123, this merge request changes the `gpart.id` and `gpart.part` anonymous union into a single `long long int id_or_neg_offset`.

If this number is positive, then it is the particle ID and the `gpart` has no hydro counterpart.

If this number is negative, then it is the offset in the `parts` array of this particle's hydro counterpart.

See merge request !118

Tidy up so that both trees use the same library dependencies