startLife tests don't silence dead incarnation's unhandled promise rejections

[michaelfig]

What is the Problem Being Solved?

In #10686

test.serial(
  expectUnhandled(2),
  'vow rejected with storable reason, upgraded, then handled',

          I suppose that the 2 unhandled rejections are an artifact of the test infra, and in real life we'd only expect 1 (the after upgrade one)? The promise created by the unhandled rejection tracker in the first incarnation will actually get GCed after starting a new life and get reported, where in a real vat, the first incarnation would be gone, and not report anything.

Originally posted by @mhofman in #10686 (comment)

Description of the Design

@mhofman asks: can we somehow silence this case in tests here (and in watch)? Maybe an integration of fake liveslots with the unhandled rejection tracker.

Security Considerations

Scaling Considerations

Test Plan

Upgrade Considerations