Proposal: takeFront and takeBack

[Roman D. Boiko]

Still, ranges which invalidate front on popFront could defer that 
(store a flag that popFront or consumeFront has been called and 
invalidate value previously returned from front in the next call 
to front). That would be intrusive with respect to such ranges, 
and a breaking change. But it would simplify client code 
significantly.

(This doesn't mean that I'm insisting, or strongly prefer some of 
my two ideas, just wanted to clearly restate the second one.)