package search & discovery

In #189 there's an open question about keywords. I've always felt that the keywords in package.json was sort of spammy -- kind of like how you could load a bunch of meta:keywords into your website in the early 2000s an like yahoo! and alta vista would be all "oh cool this page is about these words"

Additionally the search that powers npm (npms.io) is pretty centralized and relatively resource intensive to run as a service, requiring couchdb, rabbitmq and elasticsearch (and having felt some sharp corners trying to maintain it I'm loathe to do something like that if I'm not being paid).

I'm actually kind of excited that package discovery is not part of this initial manifesto as well since it leads the community to provide a bunch of different solutions.

One problem we'll face with indexing search is the same problem we'll face as humans -- finding packages. Given the distributed nature of entropic, it would be cool to see some sort of endpoint available on entropic nodes that allows someone to list the "other" entropic servers this instance is aware of.

I've got a package on my local entropic node which has dependencies on packages at `entropic.dev`, `pizza.computer`, `hypermegaglobalmart.com` and `packages.github.com` -- if I hit this endpoint it gives me this list of known registries that I can then talk to and discover their packages and what instances they are aware of.

Using this it would be much easier to write crawler to walk available nodes and determine what packages are located on each one.

As far as weighting and cataloging packages, that is still an area that needs some work, but discovery would be the first step here.