change history flushing behavior

[ckoven]

discussing history file behavior with @rgknox and @glemieux. we note that currently most of the history variables flush to zero whether or not FATES is being called for a given column. We'd like to change the behavior to give a fates_flush and a non_fates_flush value, so that on columns where fates isn't being called, the history variable is flushed to missing data, but on columns where fates is called, it flushes to zero. that way when we aggregate up to gridcell level for large-scale simulations, the history variables will still make sense for the binned quantities that revert to zero when a given bin isn't present in a given timestep.