Fix: Map/Set polyfill should store -0 in keys as +0

[teppeis]

• https://www.ecma-international.org/ecma-262/8.0/#sec-map.prototype.set
• https://www.ecma-international.org/ecma-262/8.0/#sec-set.prototype.add
• https://kangax.github.io/compat-table/es6/#test-Map_-0_key_converts_to_+0
• https://kangax.github.io/compat-table/es6/#test-Set_-0_key_converts_to_+0

[brad4d]

@teppeis are you sure this change is really needed?
I mean, did you actually see a case in running code where our polyfill for Map or Set treated +0 and -0 differently?

@shicks what do you think?

[teppeis]

@brad4d
In actual code, -0 is rarely used.
However, I think that polyfill should conform to the specification as much as possible if it doesn't have performance or other issues obviously.

[brad4d]

@teppeis what I'm saying is I suspect our code actually handles -0 correctly already even though that isn't visible in the code.
Can you show otherwise?

[teppeis]

@brad4d Current polyfill handles -0 correctly to compare with stored keys (Map#get() or #has() work well), but the stored keys are not converted to +0 (Map#keys() or #forEach() don't work).
Following tests are failed.

const map = new Map().set(-0, 'foo');
assertEquals(Infinity, 1 / map.keys().next().value);

const set = new Set([-0]);
assertEquals(Infinity, 1 / set.keys().next().value);

[brad4d]

imported and sent for internal review

[brad4d]

added a few more test cases
starting internal submission

[teppeis]

@brad4d @lauraharker thanks!