CABI mismatch for a few small cases

[lukewagner]

I found three small CABI spec discrepancies. If fixed, this test should pass. The discrepancies are commented with TODO in that test.

• When a future.{read,write} successfully completes a copy, the spec currently says that the packed i32 result code should be 0x11 (0x1 in the low nibble b/c the future is CLOSED, 0x1 in the high 28 bits b/c 1 element was copied) but it's currently 0x10.
• The spec currently says that {stream,future}.new allocates the readable end first, before the writable end, which is observable in the index allocation order.
• The spec currently says that when an async lowered calls completes eagerly without blocking, the return value is 0; currently it's 2.

I'm not especially wed to any of these details, so if there's a particular reason to change the spec, lmk!

[lukewagner]

As an update, after chatting with @dicej, we both tentatively agreed that the impl behavior is more regular, so I'll update the spec to match here.

[lukewagner]

As an update: the testcase posted above now passes on @dicej's branch \o/. However, I did find one more case (traps at the "TODO") where the current stream status is not included.

[lukewagner]

Woohoo; as of this commit all tests in the component-model tests directory pass.