RFC: Remove "unchecked" blocks

[pcwalton]

Unchecked blocks can be used to write reinterpret_cast, which can cast an unsafe function to a safe one and call it. Thus there's no real difference between an unchecked block and an unsafe block. So I propose removing unchecked from the language.

[brson]

I agree. Maybe later we can improve unsafe with additional effects, but two keywords here is extravagant.

[nikomatsakis]

+1

[catamorphism]

Agreed. The distinction between unchecked and unsafe blocks is one we don't need (we have enough confusion as it is), even if the distinction were more meaningful.

[graydon]

Agreed. There was consensus on this at the meeting, I believe. I'm assigning (tentatively) to @brson as part of the keyword-removal bug for 0.4; if it winds up taking much more work than merging the two keywords and the paths that check one vs. the other in the semantic checks, punt back to 0.5 or such. But I think this is near-term.

[brson]

Done.