Fsk issue 4691 catch bad fsu during compute moves

[pnkfelix]

Resolves third bullet of #4691: if the functional-struct-update (FSU) expression { a: b, ..s } causes s to move and s has a destructor, then the expression is illegal.

r? @nikomatsakis

[nikomatsakis]

Felix -- this looks like a good patch. My one concern is that it means that the dtor rule is enforced in two distinct places (one in moves.rs -- which is not otherwise tasked with reporting errors -- and also in gather_moves check_is_legal_to_move_from()). This concerns me lest new rules related to moves should arise. The fix I had in mind (having FSU's generate moves in gather_moves for all fields that are consumed) would seem to address that, at the cost of duplicating the logic for computing which fields are consumed. What do you think?

[nikomatsakis]

After discussion on IRC, @pnkfelix and I decided to take this patch and possibly refactor later as part of ongoing effort to stop zeroing out moved data.