reconstructions default to input datasets scale

[edyoshikun]

The output dataset's scale is wrong as it is read from the input here

Should the zyx scale be taken from the config file instead? Specifically, in the case of falcon datasets, the bug arises from here.

[edyoshikun]

The easier fix is to update-scale-metadata before starting any processing.

[talonchandler]

Closely related to #457.

Our plan is to use the input dataset's scale metadata by default, and use the config file's metadata to override the input's scale metadata.

@edyoshikun will that design work for your use case?