bug: respect method's bounding box in output

[macrintr]

To make it easier for users to reason about corgie, I think it'd be better to have the bounding box passed as an argument also match the bounding box of the output result. Right now, we snap that bounding box to the nearest CloudVolume chunks.

This is related to #28, which suggests having the info files of output CloudVolume's use bounding boxes that match the output.

[supersergiy]

This leads to hell, we will constantly need to look up the info file for the volumes we're working with. Not ever mentioning that different MIPs might have different chunk alignment, so a command might stop working after changing the processor MIP.

What's the problem with expanding to the chunk boundary? I've never encountered an issue with this in my work