a smart method to detect and jump over the offending silence is the only way to be sure of not affecting the beat

I'd like to see something which did this. The problem is that, because of the lossy way MP3s are encoded, sometimes the extra bits of silent data aren't perfectly silent. They contain some of the MP3 artifacting such as echo or pre-echo... very faint, but present. So it would be hard for an algorithm to be able to tell the difference between a gap in the encoding and a song that was "supposed to be" that way.

But clever software could still do it. Even if there was "supposed to be" a really quiet part near the end/beginning of a song, it would still be OK if the software cut off a fraction of a second's worth of it. You could have the gapless algorithm choose to only trim within a few frames of the beginning or end of the song.

It is in 3.0 but it doesn't currently offer the user such a small cross fade.

Well, consider this a feature request for that very thing, then.