van Leer smooth TVD limiter.
phi(r) = (r + |r|) / (1 + |r|) = 2·max(r,0) / (1 + |r|)
Smooth, symmetric, and achieves 2nd-order accuracy on smooth data. See van Leer (1974).