Change return type of parse

[jamiemlaw]

This PR aims to reduce the size of the library by changing the return type of the parse function. Previously, it returned a new Complex; now, it returns a [re, im] tuple. This then allows us to write things like:

const [re, im] = parse(a, b);

and use re and im as variables, rather than z['re'] and z['im'], saving quite a few bytes.

By my calculations this reduces the minified size from 9861 to 9671 bytes (a roughly 2% saving); and after GZIP, from 2890 to 2845 bytes. It also reduces the overhead of instantiating a new Complex object with each parse.

I was reticent at first to use destructuring assignment, in case it reduced browser compatibility in some way, but I see that in the hypot function, it's already used, so I don't think it'll be an issue.

Happy to hear your thoughts.

[donhatch]

My first thought is that the proposal makes it so parse() still has to do a memory allocation for the return value (a 2-element array). If optimization (or allowing future optimization) is the goal, a better signature might be to make it a member function of Complex:
Complex c;
c.parse(a, b)
That would allow a caller to reuse c for multiple parse() calls, without suffering any memory allocations at all.