Train from URLs, sitemaps, and PDFs.

New: Add one or many page URLs as training sources.
New: Train from an XML sitemap.
New: PDF sources.
Improved: All source types share the same progress display (pending, processing, done) and show page titles.
Fixed: Setup not finishing after a bot was created, and training stalling when no sources were selected.
Improved: Training is more resilient to slow sites and timeouts.