See if removing more uses of isNaN from jsexecute improves performance

[GarboMuffin]

In testing #280 it was noticed that using isNaN is a lot slower than x !== x in Firefox. There's some other parts of jsexecute where isNaN shows up. Might be worth looking into if those can be optimized.

[Tacodiva]

Have you tried using Number.isNaN() instead of isNaN()? According to the specs isNaN does type coercion where Number.isNaN() does not (isNaN("abc") returns true, where Number.isNaN("abc") returns false). I suspect the type coercion is why it's slower.

[GarboMuffin]

Using Number.isNaN does seem to have the same performance as using x !== x in #280 in Firefox