removed lib.parallel.distances

- The cython/prange-based/OpenMP versions of some of the distance calculations were
  slower and had fewer features than the C/OpenMP based ones that are now available
  with backend="OpenMP" as implemented in PR #529 so these ones (and their tests)
  are fully removed.
- removed test for deprecated core.parallel.distances
- The whole lib.parallel module was also removed because at the moment there is nothing
  in it; we can revisit the discussion about organization of lib (issue #287) when
  necessary.

package/CHANGELOG

@@ -13,10 +13,10 @@ The rules for this file:
   * release numbers follow "Semantic Versioning" http://semver.org
 
 ------------------------------------------------------------------------------
---/--/-- tyler.je.reddy, kain88-de, richardjgowers, manuel.nuno.melo, 
+--/--/15 tyler.je.reddy, kain88-de, richardjgowers, manuel.nuno.melo,
          orbeckst
 
-  * 0.xx.x
+  * 0.13.0
 
 API Changes
 
@@ -26,19 +26,26 @@ API Changes
     assumes that it is used as a container and won't give a warning if the
     number of atoms differs between frames.
   * GROWriter.fmt strings updated to use format style (Issue #494)
-
+  * removed MDAnalysis.lib.parallel.distances; use the new backend="OpenMP"
+    keyword for the functions in MDAnalysis.lib.distances (Issue #530)
+
 Enhancement
 
   * ChainReader now reports times properly summed over sub-readers (Issue #522)
   * GRO file reading approximately 50% faster for large files (Issue #212)
   * GRO file writing now will write velocities where possible (Issue #494)
   * Added bonded selection (Issue #362)
   * Spherical layer and spherical zone selections now much faster (Issue #362)
+  * new keyword "backend" for accelerated functions in MDAnalysis.lib.distances
+    to select "serial" or "OpenMP"-enabled versions of the code; the default
+    is "serial" so old code will behave as before (see Issue #530)
 
 Changes
+
   * built html doc files are no longer version controlled (Issue #491)
 
 Fixes
+
   * AMBER netcdf writer now correctly uses float32 precision (Issue #518)
   * Fixed a numpy incompatibility in `analysis.leaflet.LeafletFinder`.
     (Issue #533)

package/MDAnalysis/lib/__init__.py

@@ -21,12 +21,11 @@
 """
 
 __all__ = ['log', 'transformations', 'util', 'mdamath', 'distances',
-           'parallel', 'NeighborSearch']
+           'NeighborSearch']
 
 from . import log
 from . import transformations
 from . import util
 from . import mdamath
 from . import distances  # distances relies on mdamath
-from . import parallel
 from . import NeighborSearch