[MITgcm-support] Negative concentrations in pTracer

[Roland Young]

OK thanks (yes, typo, should have been 10^-8).

I'll change over to scheme 33 for the tracers (scheme 7 increases the 
size of the halo regions so slows it a bit, and the output looks similar 
anyway). Both those schemes still have an undershoot (same order of 
magnitude as scheme 77).

I have found a routine in the FIZHI package fizhi/fizhi_fillnegs.F which 
deals with negative tracer concentrations in what seems like a sensible 
way, so I have adapted this and I think this should be an acceptable 
solution.

Roland