add a no-op dispatch.dropExports to the kernel-to-vat pathway

[warner]

What is the Problem Being Solved?

The second preparation step for #2615 is to provide a dispatch.dropExports() call to all vats. It doesn't need to do anything yet, but as long as it exists, we can proceed with kernel-side work that might emit dropExports without fear of crashing upon a missing method.

The signature will be dispatch.dropExports(vrefs), taking an array of vat-side object references. We won't support dropping promises for now.

This needs to be added to both liveslots, the comms vat, and the vat workers (if they look closely enough at the vat/kernel protocol to notice the individual messages being sent).

Description of the Design

A later step will make dropExports do real work: it needs to delete a strong reference from the planned exports Set, and remove slotToVal/valToSlot entries for the vref, so that liveslots is no longer keeping the exported Remoteable alive. (It might be kept alive by other local references, but not liveslots).

When the kernel calls dropExports, it promises to never mention those identifiers again, at least until the vat re-exports one of them by mentioning it in some subsequent syscall. As a result, it would be an error for the kernel to dropExports the same vref twice (without an intervening re-export).

Security Considerations

Test Plan

test-liveslots.js will be augmented to invoke dropExports and make sure it doesn't crash.