force orientation of mesh to match the orientation of the geometry

keep existing vertex numbers when reading meshes (when it makes sense)

fix missing SIGN in NR

read vertex indices separated by blanks or commas in vrml (both seem to be correct)

- added CROD in BDF
- also accept small "e" and "d" in BDF floating point numbers

added Shewchuk's robust geometrical predicates (since they can be duplicated in
Tetgen/triangle I've put them in the gmsh namespace)

small fix for neg phys

also for unv

make negative physical entities behave like they're supposed to

implement revert() for all element types and use that in setVolumePositive()

- setParameter in MVertex
- revert() for quad
- minor cleanups in I/O code
- start work on overlap mesh in fourierModel

finish BDF reader for 2nd order elements

added routines output (incomplete) second order elements in BDF

SAME REMARK AS WITH UNV: PLEASE TEST THE NODE ORDERING

make reading code a bit more compact by consolidating getVertices() in a
single routine

added full support for incomplete second order elements (8-node quad,
20-node hexa, 15-node prism and 13-node pyramid):

- second order generation algorithm can now be told to generate either
  complete or incomplete elements

- MSH I/O routines can deal with all elements

- UNV I/O routines now support all the elements in the I-deas spec (I did
  my best to follow the spec for the ordering of the nodes, but I might have
  made some mistakes: PLEASE TEST IF YOU HAVE ACCESS TO A CODE OTHER THAN
  GMSH THAT CAN READ/WRITE UNV FILES)

- Use correct vertex ordering for second order simplices in UNV format
(thanks to Romuald Conty <conty@phimeca.com>)

- MLine2 -> MLine3

better fix

fix crash when adding a geo entity while a mesh is displayed

removed "primarily Delaunay" from the docs ;-)

- start work on fourierModel
- add support for pyramids in BDF (it's not in the official spec I have
  here, but all the files I found on the web use CPYRAM)

changed name of second order elements (in prevision of addition of standard
new ones like 8-node quad, 20-node hex & co)

make sure to null-terminate tmp string

generalize BDF vertex reading code to handle the special Nastran floating
point formats (-1.-1 == -1.E-1, etc.)

use atof/atoi in simple cases

- 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