Allow assert to be optimized

[geek]

If you run hoek with --trace_opt you can see that the exports.assert isn't optimized because it calls slice on arguments:

disabled optimization for 0x1c4fd5e40f49 <SharedFunctionInfo exports.assert>, reason: Bad value context for arguments value]

After changing this to manually populate msgs from the arguments the function can be optimized by the optimizing compiler.

[arb]

I'm not so sure about this change. I have no doubt this is faster and can be optimized, but there is probably a reason no one really does it this way and always uses the Array.prototype.slice.call trick.

Check out this implementation. https://github.com/petkaantonov/bluebird/wiki/Optimization-killers#32-leaking-arguments. That looks like the way to go to me personally if we are trying to squeeze optimizations.

[hueniverse]

Hmm. First, doesn't the early return already void optimization? And even if the function is not optimized, what impact does this have when the condition is true? I assume none. Are we really trying to optimize when the server is about to go boom?!

[geek]

@hueniverse from looking at the deoptimization statements... this is able to be optimized with the early return from what I saw.

It seemed like an easy win since we call assert so frequently.

[hueniverse]

Ok. Can you add a comment documenting why we are not using slice() in this case?

[geek]

@hueniverse added