variable name: b -> bundle
and other variable names:
uploadid -> upload_id, uploadobj -> upload_obj, cnt -> count
uf -> uploadfile

make import_layer explicitly take pk kwarg:
since this method is only using kwargs to get pk, we might as well extract pk
early and default it to None instead of using kwargs.get.
I actually would prefer pk to be a required arg and to dispense with **kwargs,
so we have a real signature that can check us when we make mistakes; but that
will have to depend on an analysis of where these methods are used

pass request as Bundle arg instead of setting attr after

This can help a little to make uploads more distinguishable when they don't
have different names.

also permit null file_size for UploadedData.
Sometimes an upload doesn't have a size (maybe it didn't finish uploading).
In that case, it is not really sensible to say it is 0 bytes.
It has no size.

Since some of the multifile and geopackage changes, it is no longer obvious
what should be the name of an upload. If it just contains one file, we can do
as before, but otherwise things can become more difficult.

The heuristics for this are a little complex, so they've been split off into
their own method and tests created just for this method.

No longer possible just to assume that there is one upload file type and it is
the same file type as a layer, so we need to detect the actual file type, store
it in the database and present it to users.

- add to api
- add to UploadLayer as property
- add to UploadFile as field
- add the migration to make the column
- set it from the form processing

to api, UploadLayer, set in form processing, migration to add column.

layer.name keeps filename info that is used to import, so we can't stuff actual
layer names (e.g. from inside gpkg) there without breaking import. If we change
both, any existing databases will be storing a different kind of information in
that field, which may be confusing. We can deprecate layer.name if we need to.

Whether extensions are explicitly specified in settings, or else just
the default list of extensions is used, we should consider
importers.VALID_EXTENSIONS as canon and use that in user-facing templates