Wolfram Language has an absolutely huge "standard library" though which probably accounts for some of this.
Yes, but that's explicitly part of Wolfram's design philosophy: embed as much computational knowledge as possible into the core language.
I would believe so. Things like dynamic programming (memoization) takes only a few characters, not to mention a wealth of facilities already built in. It's what I use for prototyping math-y software.
If you discount its standard library however, I think it might not necessarily beat APL.
They need to benchmark Lua. Lua is ridiculously concise when it comes to implementing algorithms.