fix: add WeakRef tracking to liveslots

[warner]

Liveslots maintains an import count: each time a vref is mentioned in an
inbound message, liveslots increments the counter. When user-level code
forgets about the Presence object, eventually (in some future turn, probably
not during one of this vat's cranks) the finalizer callback will run, and
liveslots will send a vatDecRef() message to the kernel with the old count.

This will eventually be wired up with kernel-side code that maintains a
similar counter, which will be decremented, and when it reaches zero, we'll
remove the c-list entry.

The handling of Promises needs more thought. Currently we retain a strong
reference to imported Promises until they are retired (which happens in
certain cases of resolution).

refs #1872

[warner]

I think the Zoe test failure is because the decref arrived after the vat was deleted, so kernel.js decref() threw an assertion that the vatID was bad, when instead it should have just ignored the call.

LOGGED ERROR: unknown vatID v8 [Error <Object <[Object: null prototype] {}>>: unknown vatID v8

  at Object.complain (kernel/packages/assert/src/assert.js:168:19)
  Uncaught exception in test/swingsetTests/brokenContracts/test-crashingContract.js
  at fail (kernel/packages/assert/src/assert.js:195:20)

  at Object.assert (kernel/packages/assert/src/assert.js:207:11)
  Error: unknown vatID v8
  at Object.decref (kernel/packages/SwingSet/src/kernel/kernel.js:149:12)
    at Object.complain (kernel/packages/assert/src/assert.js:168:19)
  at vatDecref (kernel/packages/SwingSet/src/kernel/vatManager/localVatManager.js:110:15)
    at fail (kernel/packages/assert/src/assert.js:195:20)
  at importDropped (kernel/packages/SwingSet/src/kernel/liveSlots.js:90:5)
    at Object.assert (kernel/packages/assert/src/assert.js:207:11)
  at FinalizationRegistry.cleanupSome (<anonymous>)]
    at Object.decref (kernel/packages/SwingSet/src/kernel/kernel.js:149:12)
    at vatDecref (kernel/packages/SwingSet/src/kernel/vatManager/localVatManager.js:110:15)
    at importDropped (kernel/packages/SwingSet/src/kernel/liveSlots.js:90:5)
    at FinalizationRegistry.cleanupSome (<anonymous>)

I'll add a more specific test and a fix.

[warner]

I'm deprioritizing this in favor of more deterministic forms of GC, we'll reassess it later.

[warner]

We're no longer pursuing this counter-based approach, instead we're relying upon the engine to give us deterministic GC so we can avoid the counters. See #2615 for the current plan.