Fix geo_interface output when bbox is None

[yellowcap]

The bbox key should not be in a geojson object if the bbox is None. If the key is present, the value is expected to be a tuple, see https://datatracker.ietf.org/doc/html/rfc7946#section-5

In ths PR, I also moved the __geo_interface__ into a mixin to reduce code duplication.

There is one design question where I was not sure what is the best:
The bbox check could be added into the dict() method directly. Depending on whether the expected output of dict is a valid geojson like dictionary, or the "full" data from the pydantic model. Please advise, I am happy to change that if its considered better to put it into the dict method.

[geospatial-jeff]

The bbox check could be added into the dict() method directly. Depending on whether the expected output of dict is a valid geojson like dictionary, or the "full" data from the pydantic model.

Personally I prefer it as written. Mutating the shape of the data on dict() may be unintuitive; but __geo_interface__ protocol is there for interoperability with other libraries (like shapely/geopandas etc.) and should follow the spec as close as possible even if that requires mutation.

The mixin works really nicely here too