fix: Clear vat table entries for promises when they are resolved

[FUDCo]

This takes care of the first part of #675, namely cleaning up the liveslots side of things.

[warner]

I don't believe we can land the two sides of the vat/kernel boundary separately: if the vat is going to forget about the p+4 identifier, the kernel must uphold its promise to never use that identifier again either.

So let's combine this with removing the kernel-side c-list entries, and unit tests to check that both have been removed. One of the tests should submit a second message with the same kernel promise identifier, and confirm that the vat receives a fresh identifier.

[FUDCo]

My memory from our meeting last week was that we concluded that the liveslots changes could stand alone, but on further reflection I think you're right: if the kernel mentions one of these promises to the vat, it seems like Bad Things would happen.

I was going for incrementalism here, but, well, never mind.