Tests shouldn't need isDarwin check with custom mkdirp

[phated]

It seems that the tests are currently passing with the check still implemented. Need to look into this.

[erikkemperman]

For what it's worth -- I thought it was quirky before the mkdirp merge...

On my mac, the isDarwin distinction was indeed necessary, or the "different modes for files and directories" test would fail. But on my mint linux box I also had to force isDarwin to be true, even though on travis (which I think is a kind of debian?) it works as intended!

I guess we may need to test the mkdirp logic with masks including the sticky/setuid/setgid bits to see what sticks (ha) on which platforms?

[phated]

I've got this mostly figured out. Just need some feedback on #185 before I commit all the changes.