Now mapDistribute structure to pull data from slaves to master and vice versa.

In case of differing pointsInstance and facesInstance it will
- read the points belonging to the facesInstance
- construct and write the mesh belonging to the facesInstance
  (so with the old, facesInstance, points)
- additionally write the current points to pointsInstance

- This avoids dependency on lex/flex and provides better encapsulation
  for buffer switching. As a result, the maximum number of open files
  only corresponds to the include depth.

- handles basic operations, references to other dictionary entries
  (with '$name' syntax) and assorted mathematical functions:

      pi(), degToRad, radToDeg, asin, acos, atan, sin, cos, tan, log,
      log10, mag, atan2, pow

The basic syntax: #calc{ ... };

NOTE the trailing ';' is required for the primitiveEntry to be
properly defined.

Now storing name of restraint or constraint in wordList in the
sixDoFRigidBodyMotion object.  Added a write method to all constraints
and restraints.

wordLists.

BUG: sixDoFRigidBodyMotion.  Fixing naming of maxConstraintIterations_
inconsistency.

foamToEnsight will merge points on coupled patches. (only for mesh, not for
patches)

leading text

* Something like ParaView_MAJOR=ParaView_3.6 would have previously
  been possible, but this violates paraview's expectation of having
  "lib/paraview-<digits>.<digits>"

* TODO? If desired, the ParaView_MAJOR could be determined
  automatically in the script.

  For example,
      ParaView_MAJOR=`echo $ParaView_VERSION | \
          sed -e 's/^\([0-9][0-9]*\.[0-9][0-9]*\).*$/\1/'`

  Or even,
      ParaView_MAJOR=${ParaView_VERSION%.[0-9]}

  Except that the second version fails if the minor version reaches
  double digits.

Option was not used anymore. Cleans up the code a bit.

- Also adjusted formatting of usage to match what argList generates

for six DoF motion bodies that are not principal axis aligned shapes
to start with.

Calculates the best match axes to give the most naturl transformation
from the Cartesian axes. The eigenvectors are returned in the order
relating to ascending magnitude of their eigenvalues - not necessarily
in a right handed triplet.

restraints when there were none read.  Was not a problem for
restraints, but constraints were stopping the code on a maxIteration
trap, as it defaults to zero.

This reverts commit edfba8be.

dictionaries - redundant.

Before two timesteps could get the same timeName and start overwriting old results.

restraint data to the classes themselves to allow class specific data
to be reported.

Have weight=7 on cells to be refined and balance. Then after refinement
there will be perfect balancing. Similar for layer addition.

It now caters for
- all surfaces having fields
- no surfaces having fields
but not a combination of those.

This directly samples on a triSurface (instead of using the isosurface of
distance=0 method)

The global sum of (integer) weights for metis&scotch has to be within the
valid range of integers. For now just divided the input (scalar) weights
with the minimum before truncating to integer.