Improve specificity of license classifiers

[xmunoz]

Our packaging ecosystem relies on a particular structured data format (classifiers) to indicate a package's legal license. However, our current system allows for ambiguity that makes some downstream data display incoherent or very difficult, and doesn't allow for some license specificity that downstream consumers need (Libraries.io and similar projects). Fixing this is a fairly small project, involving Python development, public communications, project management, and potentially a few hours of legal counsel for review.