to prevent from possible negative values under square roots

Genric function (via interface) were created for parallel function that
were had the same functionality but not the same argument types/number.

For example p_dmax and p_imax are now p_max
For example p_dsum and p_isum are now p_sum

- the velocities on one side of the culvert are taken from the flow but
  were taken always at the bottom
- now they are taken at the elevation of the culvert on that side, in
  the same way as for the tracers
- update the culvert and bergenmeersen reference files

  - new parameters in the culverts data file were added to set:
      - the angles between the inflow and outflow pipes with the
        horizontal and with the vertical;
      - the user choice regarding the automatic calculation of the
        horizontal angle based on the mesh;
  - the culvert formulation 1 is now adapted to the case of circular
    sections, like the formulation 2;
  - the results for the siphon slightly change because in the case the
    outflow is actually partially unsubmerged (which may be a small
    mistake in the original setup).

- In the previous commit only one flow direction was treated. Besides,
Pierre-Louis Ligier suggested to use the outlet's upper elevation
for the calculation of the discharge instead of the outlet's water
level in order to prevent flow rate overestimations, in particular
in the case of a falling jet at the outlet. This choice may be
discussed but is implemented here.
- The printing of the flow rates through the culverts in the listing
now only happen at the listing printout frequency.

This print was accidentally removed at revision 7842 among the
introduction of the new culvert formulation.

The formulation 1 for culverts did not fall back to the siphon
formulation. This impacts the bridge test-case of TELEMAC-2D
which now presents lower values of flow rate underneath the bridge.

[src]Algorithmic Differentiation: Merged revision(s) from branches/foxface/, including bug fixes, a modified symbolic solver and some cosmetic changes.

   * except in DREGDESIM NESTOR MASCARET

    - feature #282 of the CUE
    - the reference files for the two test-cases are unchanged

  - add the culvert modelling in TELEMAC-3D
  - add a new formulation for culverts in TELEMAC-2D and 3D, proposed and tested
    at Flanders Hydraulics (contact: Sven Smolders, sven.smolders@mow.vlaanderen.be)
  - this commit corresponds to the merge of the branch balloonfish at revision
    7837 with the trunk

[feature #347] Deallocatting array at the of telemac2d (Which includes removing all common/lu/lng now using USE DECLARATIONS_SPECIAL instead)

[all] Cleanup of the sources file : Removing tabulation transforming do label label continue in do enddo
[all] Adapting code for Fortran 2003 Standard must pass compiler option (-std=f2003 for gnu compiler)
[makefile] Moving list of files in makefile.am
[python] Adding support for section files (partel)

Telemac2D : dynamic allocation of variables for weirs and preparation of future version of weir description

This appear only when the discharge becomes equal to 0 (in such case, the value of SECT is 0) but due to relaxation, the effective discharge take in account for the computation could be different of 0 and then the computation of velocities crash.
Now, the value of SECT was saved in memory and in such case, we use the previous value. I'm not sure is we should make a relaxation on the section...