SHT timing doesn't take iterations into account

[Martin Reinecke]

I think that during the recent changes to time_SHT.c the division of the measured time by the number of iterations was lost (tdiff did this automatically, but it was replaced by a simple time difference); now the timings and GFlop reports look a bit underwhelming

[Nathanaël Schaeffer]

Yes, good catch! I pushed the fix into the master branch; new release expected soon.