Add a volume calibration scenario

[balloob]

This test scenario was identified by a cross-SDK conformance audit comparing every Sendspin client/server implementation. See the audit docs on branch claude/stream-sync-correction-sdks-AWoNC.

What it tests

Server sets volume = 50, client plays a known fixture, the harness
measures the peak amplitude of the recorded output and asserts it
matches the spec's perceptual curve.

• Conformant: ≈ 0.5^1.5 = 0.354 × original (≈ −9 dB).
• Non-conformant linear: ≈ 0.5 × original (≈ −6 dB).

Big user-experience impact, easy assertion.

Why we need it

sendspin-go and sendspin-js apply linear vol/100 gain instead
of the spec's implied (vol/100)^1.5 perceptual curve. A multi-room
group containing one go/js player and one rs/SendspinKit/cli player
will have an audible loudness mismatch at any volume below 100.
Without a test, this regresses silently — and the SDKs that delegate
to platform audio (sendspin-cpp, sendspin-jvm) can't verify their
platform is doing the right thing.

See spec issue: volume curve for the proposed normative text.

Implementation sketch

• Existing audio-hash machinery, comparing peak amplitudes between
  volume = 100 and volume = 50 runs of the same fixture.
• Conformant range: 0.30–0.40 (allowing for window/edge effects).
• Failing range (linear): 0.45–0.55.

---

Source: docs/volume-curve.md §recommendations