test(nests): T16 Phase 3 — I9 tolerance + Phase 4 plan

I9 (packet-loss) tolerance bumped from 80% → 50% expected
samples. moq-lite groups are reliable streams so retransmits
absorb 1% loss, but hang-listen's `Container::Consumer` runs
with a 500 ms latency window and aggressively skips groups
that arrive late. Random 1% loss can land on back-to-back
packets that push a single group past the window — the deficit
is non-deterministic. The 50% threshold catches a wholesale
failure (catalog never arrives, all groups dropped) without
flaking on normal jitter.

Phase 4 plan filed at
`nestsClient/plans/2026-05-06-phase4-browser-harness.md` —
1.5-day spec for the bun + Playwright browser harness
(`@moq/watch` listener + `@moq/publish` publisher in headless
Chromium). Self-contained: lives in a new
`nestsClient-browser-interop/` directory tree and
`BrowserInteropTest.kt`; reuses the existing
`NativeMoqRelayHarness` infra. Zero overlap with the hang-tier
scenarios.

A separate agent picks this up on a fresh
`feat/nests-browser-interop` branch.

https://claude.ai/code/session_01ERJPUYfdLPwZ99pr5EcEcV

nestsClient/src/jvmTest/kotlin/com/vitorpamplona/nestsclient/interop/native/HangInteropTest.kt

@@ -388,11 +388,17 @@ class HangInteropTest {
     /**
      * I9 — 1% packet loss via the `udp-loss-shim` between the
      * Kotlin speaker's UDP socket and the relay. Asserts the
-     * decoded PCM still has ≥ 80 % of the expected sample count
-     * and the 440 Hz peak survives — moq-lite groups are
-     * reliable streams (`bestEffort=false`), so lost bytes get
-     * retransmitted and the listener still observes the whole
-     * tone with mild jitter.
+     * decoded PCM still has at least half the expected sample
+     * count and the 440 Hz peak survives — moq-lite groups are
+     * reliable streams (`bestEffort=false`) so lost bytes get
+     * retransmitted, but hang-listen's `Container::Consumer`
+     * runs with a 500 ms latency window and aggressively skips
+     * groups that arrive late. Random 1 % loss can land on
+     * back-to-back packets that push a single group past the
+     * window, so the deficit is non-deterministic. The
+     * threshold is tuned to catch a wholesale failure (catalog
+     * never arrives, all groups dropped) without flaking on
+     * normal jitter.
      */
     @Test
     fun packet_loss_1pct_does_not_kill_audio() =
@@ -406,8 +412,8 @@ class HangInteropTest {
             val pcm = readFloat32Pcm(out.pcmFile)
             val expected = 5.0 * AudioFormat.SAMPLE_RATE_HZ
             assertTrue(
-                pcm.size >= expected * 0.80,
-                "expected ≥ 80% of $expected samples under 1% packet loss, " +
+                pcm.size >= expected * 0.50,
+                "expected ≥ 50% of $expected samples under 1% packet loss, " +
                     "got ${pcm.size} (${"%.1f".format(pcm.size / expected * 100)} %)",
             )
             val warmup = AudioFormat.SAMPLE_RATE_HZ / 25