2) Any lossy re-encoding step degrades quality. You'll get the highest quality by saving the stream direct from the server and skipping the re-encode step. Just pass --extract-audio and skip all those other arguments.
http://listening-test.coresv.net/results.htm
The other kind of comparison to consider is transparency - the ability of a lossy codec to be indistinguishable from the uncompressed source. MP3 is supposed to achieve this at 192 kbps, while OPUS can do it in 128.
https://wiki.hydrogenaud.io/index.php?title=Transparency
The best choice, at the end of the day, is wholly dependent of the context of the usage. If all of the source is in MP3 to begin with, it doesn't make sense to convert.
AAC has very wide support. Only convert AAC to MP3 if your player really doesn't support AAC.