Implement skips for stringify array format comma

[decimoseptimo]

Fixes #302

[sindresorhus]

Is this PR done and ready to be re-reviewed?

[decimoseptimo]

Yes it is.

[DV8FromTheWorld]

This is a breaking change to the default functionality of stringify for comma and separator array modes given that skipNull and skipEmptyString are default false but the original functionality of stringify for these 2 modes behaved as though skipNull and skipEmptyString were true.

Anyone relying on the original behavior will now have different results after this change.

Doesn't this require a major version revision?

[decimoseptimo]

The documented default behaviour for skipNull and skipEmptyString is and had been false. Because stringify was broken, it wasn't checking for those, and it would resemble to true.
It should at least added to the release notes that it would unbreak things, it's up to the owner about the version bump.

[DV8FromTheWorld]

The documented default behavior of skipNull and skipEmptyString has always been false, yes. But you've modified the default behavior of the comma and separator methods which weren't following those options before. It wasn't "broken" per-se, it was just not respecting the options. The functionality of comma has existed for 2 years and predates both of these options, so it isn't really correct to say it was broken.

The point I'm making is that this change breaks assumptions that already exist in the library that the comma and separator options always stringify by dropping any falsy values. Any software that consumes this library and that is built on that assumption is going to encounter problems if this change is released under a non-major-version change considering that most library inclusion into projects uses fuzzy matching based on minor+revision.

I'd suggest releasing this change as a major revision to 7.x given the breaking nature of it.