Fix potential bugs if no initialisation to 0 is done for work arrays as done in 3D

In #982 (closed), @michael.turnbull and @thomas.benson found configurations where bugs appear in parallel. The fixes were to initialise a few work arrays to 0 before calling VECTOR with a specific mask (not the more usual one when OPTION FOR THE TREATMENT OF TIDAL FLATS = 2 (i.e. dry elements frozen) or ELEMENTS MASKED BY USER = YES), e.g. formulation = FLUBOR.

The same should also be done in TELEMAC-2D.