Improved lume.distance()'s performance on non-JIT

Replaced use of the exponent operator with multiplications, yielding a
performance increase on non-JIT Lua:

[5000000 calls, Lua 5.1]
old func: 2.03 seconds
new func: 1.17 seconds

[5000000 calls, Lua 5.2]
old func: 1.60 seconds
new func: 0.89 seconds

[2000000000 calls, LuaJIT 2.0.2]
old func: 0.89 seconds
new func: 0.89 seconds

lume.lua

@@ -43,7 +43,9 @@ end
 
 
 function lume.distance(x1, y1, x2, y2, squared)
-  local s = (x1 - x2) ^ 2 + (y1 - y2) ^ 2
+  local dx = x1 - x2
+  local dy = y1 - y2
+  local s = dx * dx + dy * dy
   return squared and s or math.sqrt(s)
 end