Towards `Extractor` output format

[PeterKraus]

We should think about how extractors can report their desired formats at 
some point (thinking e.g. if we'll have issues with pandas 3 etc)

Originally posted by @ml-evs in datatractor/beam#3 (review)

[PeterKraus]

What we have now:

• Support for specifying {{ output_type }} in templating, but requires the extractor to support it too,
• Support for specifying supported_output_filetypes for an Extractor, at the moment it's "global" (i.e. not per-usage), but it could be adapted to match supported_filetypes

Of course, this deals with FileTypes only and not any returned in-memory objects. For usage.method == python, it might be sensible to also provide something like required_formats that indicates what package(s) (e.g. pandas, numpy < 2.0 etc.) needs to be installed in the parent env to be able to handle the returned object.