Add rich reprs to STAC objects

[TomAugspurger]

IPython / jupyter define a protocol where objects defining methods like _repr_html_ will have the ability to produce rich representations in a notebook. See https://ipython.readthedocs.io/en/stable/config/integrating.html.

I think this would be very useful for exploring STAC objects. dask/distributed#4857 has some designs / code that might be worth stealing. In particular the handling of nested objects would apply to catalogs, collections, and even items.

See http://matthewrocklin.com/blog/2019/07/04/html-repr for some background.

[DahnJ]

This seems like an immensely useful feature to have, especially given the nested structure STAC has.

PR #573 was closed since it used an extra dependency jinja2. The agreed-upon solution was to create a new package pystac-jupyter. Did that ever get made? If not, what would be needed to help it get made?

In case it didn't get made: would it perhaps be possible/convenient to only include jinja2 as an optional dependency and only attach/display the HTML representations if jinja2 is installed? I haven't thought much about this, just thought I'd mention it.

cc: @moradology

[TomAugspurger]

IMO, this would have to go in pystac itself to be worthwhile. https://ipython.readthedocs.io/en/stable/config/integrating.html#formatters-for-third-party-types mentions a way for third-party libraries to register formatters for types in other libraries, but that's not too user-friendly.

I think an optional dependency is a good compromise. The _repr_html_ would then just need a

try:
    import jinja2
except ImportError:
    return repr(self)
...

And in case it's helpful, rich.inspect can give a (verbose) repr:

But I think that the nested nature of STAC objects means things like HTML dropdowns / disclosure triangles will be helpful.

[duckontheweb]

Closed by #743