[ie/facebook] New approach extractor

- extract data from JSON strings with regex instead of traverse_obj in dicts
- more lenient url pattern
- extract linked external video
- improve thumbnail extraction
- get metadata from SD & HD media streams if ffprobe is available
Known bug: videos in comments may be included