cleanup high order

   1) this removes the asymmetry in all the routines involving mesh
      elements associated with points

   2) it makes it possible to use point-based datasets in the
      model-based post-processing module

  **********************************************************
    PLEASE TEST --- ESPECIALLY THE INPUT/OUTPUT ROUTINES!
  **********************************************************

* made GModel::getEntities() more efficient and generalized its use

* implemented first rough draft of VTK mesh writer

******************************
*   WARNING - PLEASE TEST    *
******************************

- changed I/O so that we NEVER change vertex number ids once a mesh has
been generated (or loaded)

- removed dataIndex from MVertex: stepData in PVewDataGModel now directly
uses the vertex/element number ids. This can lead to increased memory
use in corner cases (for which PViewDataList works fine, so it's not a
big deal), but it is simpler and faster

untabify

A minimal Gmsh "null" model can now be embedded with no external dependencies

fix BDF small field export on Windows (apparently Windows does not conform
to ANSI C and uses 3 digits in %e instead of 2)

cleanup previous commit + move "carve hole" routines in separate file

- only save mesh vertices that are used by elements

- fix warnings + indent + memory leaks in new Field classes

- temporary fix to avoid crashes in new field manager when post-pro
  views are reloaded or removed at run-time

- tentative fix for the "mysterious" crashes on Windows:

  I think that the problem comes from the new recursive tetra
  classification, where we probably overflow the stack on Windows. The
  fix is to request a larger initial stack size through the linker.

- added option to control label saving in unrolled geo files

- only test for neg jacobians for 2D/plane meshes for now

- fixed several compiler warnings

- preliminary work to interface more parser functions with general
  GEntities

- we can now export simple GModels as.geo files (points, straight line
  segments, plane surfaces, ruled surfaces and volumes are treated
  correctly; all non-straight-line curves are exported as splines; all
  non-plane/ruled surfaces are not exported)

- tentative fix for the unv export problem reported on the mailing
  list (fprintf() buggy on some machines??)

- new IO (both read and write) for structured grids using the NASA
  Plot3D format (for Olivier)

- finished Nastran BDF implementation (read+write in all BDF formats:
  free field, small field & long field)
- improved UNV and Medit import
- added detection of bad vertex indices in all import routines
- fixed bug in 2nd order quad constructor

- changed UNV output to match MSH output w.r.t. physical groups
- new UNV file reader

Added support (both read and write) for the Nastran Bulk Data File
format (BDF).

This might need some adjustments depending on what the actual spec
is: I just reverse engineered some BDF files I found on the web :-)

Implemented new binary MSH file format (both read and write).

A 250Mb mesh file (6 million tetrahedra) is now read or written to disk in
less than 10 seconds (on my powerbook's slow hard drive).