heap: align > MIN_ALIGN can trigger undefined behavior

[thestinger]

The non-standard jemalloc API makes dealing with this into the caller's problem.

[steveklabnik]

Triage: we have recently updated to jemallc 4.0.0, but we still may want to worry about this.

[steveklabnik]

Triage: not aware of any changes here. Tagging as "unsound" and for the libs team, as this would be their issue.

[Mark-Simulacrum]

Nominating for discussion by libs team since this is apparently unsound; so we should at least get eyes on it.

[alexcrichton]

Discussed during libs triage today, this is mostly relevant for #32838 now and I've left a comment on the associated PR explaining how this is relevant. This will likely get closed in time as the allocator trait lands, stabilizes, and we audit various locations in libstd.

[alexcrichton]

This was handled as part of #42313, so closing.