Fix for problem with HLLE and near-dry states

[carlosmunozmoncayo]

When running a 1D dam-break simulation with a initial near-dry state, the HLLE solver gave a wrong solution. The following image corresponds to the output of hl=1, hr=1e-33, ul=ur=0, using the Classic solver with mx=4000 and order=1:

This seems to be caused by the denominator (s1-s2) getting close to zero while computing middle states. The proposed fix just sets the speeds and waves as zero whenever both hl and hr are smaller than a tolerance (1e-14). This is the output with the same initial values after applying the changes:

[mandli]

Good suggestion! This solver was really never meant to really handle dry-states but it should be able to handle simple ones.