ENH: OSspecific - softlink handling (fixes #164)

Links are followed in most cases, with some notable exceptions:

- mv, mvBak:
  renames the link, not the underlying file/directory

- rmDir:
  remove the symlink to a directory, does not recurse into the
  underlying directory

See merge request !51

Fixes #192

so will be out of date (w.r.t bounding box, subdivisions) when the mesh moves.
Only when all cells stays in all the same boxes can you skip rebuilding it
so this was not deemed worthwhile. Fixes #172

STYLE/ENH: reduce code duplication for coded boundary conditions (issue #184)

- relocate common dictionary output into codedBase class

@Mattijs please test with a few examples as well

See merge request !54

- remove old VTK_CONVEX_POINT_SET code, since VTK_POLYHEDRON exists
  since several years

ENH: improve robustness of paraFoam script

- only check the relevant plugin types,
  fallback to native reader if needed/possible.

- no engrid development since 2013

- return 0 if not under git, since nothing particular can be said
  about the build number.

- explicitly define which git-dir is to be used.
  This ensure we don't accidentally get some values from some unrelated
  project in the parent directory.

- explicitly use List<int> instead List<label> for API compatibility,
  even when 64-bit labels are in use.

- many systems may not have boost or FFTW installed, or in an older
  version. Using ThirdParty for them should lead to a more robust
  build process.

- triggered by dash.

- the uncollated version becomes quite difficult to process.

  Caveat: cannot use "collateTimes true" for non-static geometries

- conflicts resolved:
    src/conversion/ensight/file/ensightGeoFile.C

- regression introduced by ef9a14b2

- similar to foamToEnsightParts, provide a data/ subdirectory to hold
  the time-varying field information (collated output only).

- widen data mask to avoid possible overflows with many time steps

- ensight does supports geometry in a separate file, but for the
  surface writer we claim that this is not true. This allows
  the field writers to control where the mesh/case files land.

  With 'true', the uncollated output would incidentally create
  things like this:

  Previous (w/ fields)        | Now (w/ fields)
  ----------------------------+----------------------
    0.05/surf.case            |
    0.05/surf.mesh            |
    0.05/p/surf.case          |  0.05/p/surf.case
    0.05/p/surf.mesh          |  0.05/p/surf.mesh
    0.05/p/surf.0000.p        |  0.05/p/surf.0000.p

  Previous (w/o fields)       | Now (w/o fields)
  ----------------------------+----------------------
    0.05/surf.case            | 0.05/surf.mesh
    0.05/surf.case            | 0.05/surf.mesh

  Move the separateGeometry() flag to the *.C code, where it is less
  likely to be mistakenly altered.

Feature noise

New functionality includes:
- run-time selectable noise models: point|surface
- run-time selectable window models: Hanning (+ options symmetric, extended), uniform
- calculates PSD (Pa^2/Hz) and dB/HZ; SPL (Pa^2) and dB
- calculates 1/3 octave data, with centre frequency 1kHz

surfaceNoise only:
- reads ascii/binary ensight surface data (requires collateTimes option)
- generates graphs for surface average quantities
- operates in parallel

See merge request !50