Fix compilation issue with Intel compiler

When compiling gmsh 4.0.0 with Intel 2017 and 2018, the following error message is encountered:

/home/mrasquin/software/gmsh/src-gitlab/Mesh/QuadTriUtils.cpp(381): internal error: conv_class_prvalue_operand_to_lvalue: couldn't convert to ptr (!t) ? internal_tri.insert(i) : internal_quad.insert(i);

Not sure why Intel complains about this rvalue to lvalue conversion in this specific case while gcc compiles just fine. A quick fix implemented in this merge request consists in replacing (!t) ? internal_tri.insert(i) : internal_quad.insert(i); and similar instructions in the same file by their if/else counterpart, for instance: if (!t) internal_tri.insert(i); else internal_quad.insert(i);

This implementation is less compact but at least the Intel compiler runs fine with this change. Please let me know what you think.