Fix a bug that would prevent global_min_speed in the multidomain log file from exceeding zero, if there really is nonzero velocity everywhere.
Improve comments, including the possibility of differences between the log statistics and results in domain%max_U and the netcdf grid outputs.
The main issue is that the log-statistics are generally computed immediately AFTER flux correction has been applied, whereas the max_U statistics are updated after every domain step, but before parallel communication and flux correction. Arguably neither is more "correct", but there is potential for flux correction to change the max-stage (or max-speed) at the site of the correction, and if these are the domain-specific extremes they could show up in the log statistcs.
Minor edits related to log file statistics:
global_min_speed
in the multidomain log file from exceeding zero, if there really is nonzero velocity everywhere.