Removing copies causing segfaults

[alexcrichton]

When transitioning from oldmap::HashMap to LinearMap in #5523, I ran into a number of spots which caused segfaults when compiling some time between stage1 and stage2.

Each segfault could be fixed with a copy, although the copy isn't necessary in most cases. This mostly showed up with types of the form &*T, and the copy ended up yielding a *T type.

I wasn't able to track it down much further, but there are a number of spots throughout the compiler marked with a FIXME for this issue.

[nikomatsakis]

I am 99% sure I know the origin of this problem. As part of my fix for #5910, I encountered several of these copies at locations where hashtables were being mutated while being iterated. This was going undetected due to numerous bugs in the @mut write guard code. Therefore, I am going to close this bug as a dup of #5910.