Lifetimes are unsuited as write lock identifiers

[RalfJung]

Analyzing the test failure in hashmap.rs revealed we have a problem with lifetime-based subtyping.
In insert_hashed_nocheck, _9 has type core::option::Option<collections::hash::map::Entry<ReScope(Remainder(BlockRemainder { block: NodeId(9681), first_statement_index: 0 })), K, V>> but then we have _17 = ((((_9 as Some).0 as Vacant).0, and _24 = _17, and _17 has type collections::hash::map::VacantEntry<ReScope(Misc(NodeId(9719))), K, V>. The lifetime changed.

[RalfJung]

Actually it turns out that this is the same problem as the more fundamental "we can't identify write locks by their lifetime". Which is a problem in the model, not the implementation.