[v3.0.0bx] Support learning URLs

[dlqqq]

Problem

Right now, /learn is limited to learning local files. It would be nice if we removed the need for locally downloading knowledge data and instead implemented support for /learn to directly accept a URL.

Proposed Solution

Implement web crawling of URLs passed to /learn.

[JasonWeill]

By "web crawling," do you mean crawling a single page, or (optionally) crawling pages linked from a specified URL as well?