RPL mini-challenge: Create an anti-Identity Matrix

[DavidM]

These are some great solutions so far, please keep them coming!

I should have foreseen the exact vs. approximate number issue and clarified it before this came up, but I think the appropriate standard in this case is to assume approximate mode for all constants and calculations. Though there are obviously exceptions, I still think that should help more implementations than it hurts.

I'm open-minded on the topic, though. If the majority feel otherwise, please speak up.