Issue with `time.enforce_checkpoint_time_dt`

[rthedin]

When running with time.enforce_checkpoint_time_dt set to true, we get negative CFL numbers. Everything seems to be working fine, though, but the reported CFL is weird. That happened even when the requested checkpoint interval was very far away from the beginning of the simulation.

Note that I'm using time.checkpoint_time_interval, and not time.checkpoint_interval if that matters. I also have time.initial_dt set. However, commenting the time.enforce_checkpoint_time_dt out solves it.

Example output:

Begin simulation:
  Run till 20000 seconds
  Adaptive timestepping with max. CFL = 0.5

==============================================================================
Step: 1 dt: 1 Time: 0 to 1
CFL: -0.326841 (conv: -0.199242 diff: -0 src: -0.204217 )

Godunov:
  System                     Iters      Initial residual        Final residual
  ----------------------------------------------------------------------------
  MAC_projection                 1       0.0003827600943       1.496445838e-05
  temperature_solve              1        0.001360573334       2.273736754e-13
  velocity_solve                 1         0.03196227125       1.337351727e-05
  Nodal_projection               1        0.002148186039       6.323805592e-06

WallClockTime in Evolve() for step 1: Pre: 0.00477 Solve: 1.761 Post: 0.0117 Total: 1.778
Cumulative WallClockTime in Evolve(): 1.778

==============================================================================
Step: 2 dt: 1 Time: 1 to 2
CFL: -0.326842 (conv: -0.199243 diff: -0 src: -0.204217 )

Godunov:
<...>