[TetraIntersec] Several bug fixes in Grandy's implementation
+ triple product inconsistency was not properly detected (now using original deltas from douple prod computation)
+ f/F factor set to 20 (not 500) as in Grandy's article -> (will be set to 100, see next commit)
+ epsilonEqual used when necessary: when it provides more points in polygon A or B (risk is completely missing a point in polygon!)
+ better handling of degenerated case where PQR triangle is in h=0
plane, or when P,Q or R close to a tet corner
+ fixed surface-edge intersection test: triple product equality (and zeroing) must be checked with care
+ using 'long double' is not necessary -> double are enough
