Support URL files with up to millions of lines

[adutra]

This came up while reviewing #399: some users are using giant urlfiles with millions of URLs inside.

This file size isn't how urlfiles were designed to work: indeed currently when a urlfile is parsed, all the parsed URL instances are stored in memory. See AbstractFileBasedConnector#loadURLs.

We should modify that method to return a Flux instead, and merge it with other fluxes.

┆Issue is synchronized with this Jira Task by Unito