Arrival times not correct in regions of subsidence

[rjleveque]

Arrival times are computed src/2d/shallow/fgmax_frompatch.f90 in a way that assumes the sea level and initial shoreline did not change due to subsidence or uplift during an earthquake, which is often not true in the nearfield. For example, at an onshore point where the initial topo B0 > 0 but after subsidence B < 0, the arrival time should be when this point first gets wet. But the current test checks when h + B > arrival_tol which isn't until the water has gone up to depth h > B0 - B + arrival_tol. This may not be easy to fix since B0 is unknown at this point and this grid patch may have first appeared only after the topography motion stopped.