Use offline registry, and handle new npm quatting rules

[laggingreflex]

How about using an offline registry, like all-the-package-names?

Using offline registry will also enable checking package names according to new package moniker rules, otherwise it'd be 59,049 http requests for jsonstream.

I implemented this in a PR to npm-name sindresorhus/npm-name#27, but thinking now, I think it might be better suited to implement in this repo.

Thoughts?

Edit: released my own package for now: name-exists

[sholladay]

We definitely do need some way to figure out the canonical package name to use when communicating with the registry. Under the new npm rules, package names that don't technically exist (and therefor would not be considered a squatter) could be unavailable for use due to similarly named packages with different punctuation.

Ideally, the registry APIs would normalize package names on our behalf, but since they don't for legacy reasons, using offline registry data from all-the-package-names to scan for what the canonical name is upfront seems like a reasonable idea.

There are three places where squatter fetches data over the network using the package name:

1. For the heuristic in significant-downloads.js.
2. For the heuristic in has-binary-or-dependent.js.
3. For the full metadata in index.js

As an aside, there is offline data for download-counts and dependent-counts, but the data is quite old. In the future, we could consider using those as fallbacks.