Fuzz bug with atomic wait #6450

The fuzzer produced this delightful module:

(module
 (type $0 (func))
 (memory $0 23 256 shared)
 (export "wait_and_log" (func $0))
 (func $0
  (drop
   (memory.atomic.wait64
    (memory.atomic.wait64
     (i32.const 0)
     (i64.const 0)
     (i64.const 0)
    )
    (i64.const 0)
    (i64.const 0)
   )
  )
 )
)

I haven't had time to investigate in detail, but our interpreter returns without a problem while V8 thows RuntimeError: operation does not support unaligned accesses. I'm not convinced that this program actually does any unaligned accesses, so this might be a v8 bug, but I also noticed that our interpreter doesn't seem to check alignment for atomic operations (unless I missed it).