Relax tolerance for truncating small snocan values in CanopyFluxes

The default of 1e-13 was letting through some state updates that look
like they should have been truncated to 0: each time step, snocan was
decreasing by about 12 orders of magnitude.

Addresses #2444

src/biogeophys/CanopyFluxesMod.F90

@@ -1624,7 +1624,8 @@ subroutine CanopyFluxes(bounds,  num_exposedvegp, filter_exposedvegp,
       ! snocan < rel_epsilon * snocan_baseline will be set to zero
       ! See NumericsMod for rel_epsilon value
       call truncate_small_values(fn, filterp, begp, endp, &
-         snocan_baseline(begp:endp), snocan(begp:endp))
+         snocan_baseline(begp:endp), snocan(begp:endp), &
+         custom_rel_epsilon=1.e-10_r8)
 
       if ( use_fates ) then