Report phase and observed values from native standard I/O regression failures

[LunaStev]

Observed at PR #631 head d811e0c2accc8997eecb5071f8752f88e9fb44e4. Failing CI job.

The new native I/O harness currently reports only 12 != 0, 11 != 0, 16 != 0, or the executable name with a 10-second TimeoutExpired. run_fixture includes stdout/stderr, but the Wave fixtures emit no relevant values. Multiple distinct predicates share exit codes.

This blocks diagnosis across the four failed jobs in #631: capture exit 12 hides child exit/payload/count; capture-close exit 11 hides wait result versus signal termination; timed-send exit 16 hides the returned count, and a hang could occur before or after the timed call.

Bounded change: add concise phase/value diagnostics to these fixtures and include the fixture name, target, actual return code and bounded captured output in harness failures. Keep diagnostics away from stdout payload channels that tests intentionally capture. Preserve existing exit checks and process-tree timeouts.

Acceptance:

• Failed capture checks identify raw wait status, wait return, count and expected/observed bytes as relevant.
• Timed socket failures identify operation phase, count, error kind/native code and elapsed time.
• A timeout preserves the last emitted phase without depending on buffered stdout flushing.
• A small harness test demonstrates nonzero and timeout diagnostics without requiring a native platform.

This concerns tests/fixtures/io and tools/test_std_io_runtime.py; #576 separately covers tools/run_tests.py case JSON. It is an observability task, not a substitute for resolving the runtime failures.

Source references:

• tools/test_std_io_runtime.py
• tests/fixtures/io/tcp_timeout.wave
• tests/fixtures/io/capture.wave