Attempt to convert the main output variable from the BMI model

[hellkite500]

Currently the model engine assumes main output variables are represented as m/timestep.
Each timestep, this response variable is converted to m^3/s based on the feature's area, and then given to the nexus (here).

A serious unit problem exists if the main output variable in the BMI model is NOT provided in a units of m. The model engine should probably attempt to convert the response variable it gets here to m for internal use at this point, and if the unit is unconvertible in this case, it should likely terminate at this point with a prominent error.

[hellkite500]

@seyounger thanks for bringing this up!

[PhilMiller]

The Raytheon team is actually working on exactly these sorts of concerns at the moment. I'll make sure this is included.

[robertbartel]

There is also at least the potential for another problem if simulation_time.get_output_interval_seconds() (which is what gets pass to get_response for timestep size) isn't actually 3600, even if the response was in meters.

Also, strictly speaking this applies to any implementation of HY_CatchmentRealization. We're used to dealing with BMI formulations because that's all that exists at the moment, but it may not always be so. That said, a conversion solution is probably easily applied generally.