Paper fixes

Latest version of the paper, still need to add some numbers and update the scaling and task plots.

See merge request !7

fixes to get this to work with the latest ompss. have to abort at the end, otherwise nanos++ hangs. can't figure out why at the moment.

add libcblas, needed for the dgemm function. the header file is tested for in the config process, but the library was never linked.

Autotools update

Bring autotools use into line with SWIFT. Adds various configure options (sanitizer, debug,
optimization) and removes remaining compiler hardcoding. Also adds a version file and
fixes a problem building on COSMA with CBLAS and LAPACK.

See merge request !4

removes compiler hard coding, adds configure option for controlling optimisation, debugging and sanitizer

build system for doc/ and remove all generated files and symbolic links.
Also private headers so that the distribution tarball builds.
Use gnu standards (COPYING, README, NEWS, ChangeLog) globally, no need to
specify in each Makefile.

Also need to add other generated files that should not be committed, namely autoconf.

Bugfix

Fixes issue #4 reported by Matthieu and another bug found by Aidan.

See merge request !2

fixed bug in single-precision iact_pair_direct, was wrong masses in local parts copy. also added matthieu's fix for issue #4 and cleaned-up some of the ICHECK stuff.

Merge from svn

Merge-in the changes I made to speed things up a bit in the sorted interactions. Was committed to SVN first because I missed Peter's mail regarding the new repository, using it to practice using git now ;)

See merge request !1

merged-in changes made in the svn last night, i.e. added the single-precision code from test_bh.c and cleaned up the multipole interaction computation a bit.

Added the calculation of the orthogonal distance to the diagonal linking the cells for each particles. Should help understand the bias.

Added a new example that solves Poisson's equation using the fast multipole method instead of the normal Barnes-Hut algorithm

Updated version of the multipole calculation where the matrix is computed on the fly. Algorithm is now O(N).