Set nfire in every time step

This shouldn't have any impact on the evolution of the model, but it
seems needed so that the nfire diagnostic field is correct: Without
this, nfire seems to persist from one time step to the next in
conditions where it would remain unset and should in fact become 0; this
is relevant in transient landcover cases.

Partially resolves #31

src/biogeochem/CNFireLi2014Mod.F90

@@ -557,6 +557,7 @@ subroutine CNFireArea (this, bounds, num_soilc, filter_soilc, num_soilp, filter_
         c = filter_soilc(fc)
         g = col%gridcell(c)
         hdmlf=this%forc_hdm(g)
+        nfire(c) = 0._r8
         if( cropf_col(c)  <  1.0 )then
            if (trotr1_col(c)+trotr2_col(c)>0.6_r8) then
               farea_burned(c)=min(1.0_r8,baf_crop(c)+baf_peatf(c))

src/biogeochem/CNFireLi2016Mod.F90

@@ -579,6 +579,7 @@ subroutine CNFireArea (this, bounds, num_soilc, filter_soilc, num_soilp, filter_
         c = filter_soilc(fc)
         g = col%gridcell(c)
         hdmlf=this%forc_hdm(g)
+        nfire(c) = 0._r8
         if( cropf_col(c)  <  1._r8 )then
            fuelc(c) = totlitc(c)+totvegc(c)-rootc_col(c)-fuelc_crop(c)*cropf_col(c)
            if (spinup_state == 2) then