allows us more flexibility in the reprojection workflow to choose between time and memory usage, i.e. instead of writing out an intermediary WorkUnit between the ic->wu generation and reprojection we can just pass in our already loaded WorkUnit but have the output be a sharded WorkUnit, using more memory but potentially saving time.
new write_output flag in reproject_work_unit to create write the output of reprojection instead of re-materializing as a WorkUnit in memory
also integrates the reproject_lazy_work_unit function into the main reproject_work_unit function.
resolves #678
allows us more flexibility in the reprojection workflow to choose between time and memory usage, i.e. instead of writing out an intermediary
WorkUnitbetween the ic->wu generation and reprojection we can just pass in our already loadedWorkUnitbut have the output be a shardedWorkUnit, using more memory but potentially saving time.write_outputflag inreproject_work_unitto create write the output of reprojection instead of re-materializing as aWorkUnitin memoryreproject_lazy_work_unitfunction into the mainreproject_work_unitfunction.