[Exploratory] Large data (and potentially other such things) as packages

Not sure if/how/what we'll formalize into CEP(s). But I'm starting to think about how we could use `conda` to deliver large data sets and other "non-code" binary blobs. Think reference genomes for `bioconda` packages, corpora for `nltk`, pre-trained models for `${your favorite new gen AI package}`, etc.

We can currently do things like just generating "huge" (multi-GB) packages or (ab)using post-link/activation scripts to run `wget some-suspect-url`, but I'm wondering if, as a community, we can come up with more clever solutions.