voxtral + kyutai-tts: fix wrong image tag / wrong endpoint paths; fish-s2: env-selectable model variant

Three fixes from the second-wave deploy attempts:

* voxtral: vllm/vllm-omni doesn't publish a `latest` tag — pull
  failed with "manifest unknown". Pinned VOXTRAL_VLLM_TAG to v0.18.0
  (released 2026-03-29, the day after the Voxtral 4B TTS release —
  first cut with Voxtral support).

* kyutai-tts: NillPointer wrapper exposes ONLY /health (root) and
  POST /v1/audio/speech. No /v1/models, no /v1/audio/voices —
  those return 404. Verified by /openapi.json against the live
  container. Compose healthcheck + playbook wait + verify steps
  all repointed at the actual paths. POST /v1/audio/speech is now
  smoke-tested with a RIFF WAV assertion (same pattern as fish-s2).

* fish-s2: added FISH_S2_MODEL env var so the model variant is
  swappable via .env without rebuilding. Both s2-pro (default) and
  s1-mini are pre-pulled into the bind-mount; LLAMA_CHECKPOINT_PATH
  + DECODER_CHECKPOINT_PATH now use ${FISH_S2_MODEL:-s2-pro}.
  s1-mini was originally gated on fishaudio's HF org (401), but
  niobures/OpenAudio-S1 mirrors the same files openly — pulled
  from there via a one-shot snapshot_download.

stacks/fish-s2/compose.yaml

@@ -63,6 +63,13 @@ services:
       # checkpoint.
       - COMPILE=${FISH_S2_COMPILE:-1}
       - API_PORT=8080
+      # Model selection. Override Fish's Dockerfile defaults so we can
+      # swap variants via .env without rebuilding. Both checkpoints are
+      # pre-pulled by the deploy playbook into bind-mounted checkpoints/.
+      #   s2-pro   — 4B class, ~17 GB VRAM, slow but highest quality
+      #   s1-mini  — lighter, ~3-5× smaller model.pth, much faster
+      - LLAMA_CHECKPOINT_PATH=checkpoints/${FISH_S2_MODEL:-s2-pro}
+      - DECODER_CHECKPOINT_PATH=checkpoints/${FISH_S2_MODEL:-s2-pro}/codec.pth
       # Hugging Face cache for model weights — first start pulls
       # fishaudio/s2-pro (~9 GB BF16) into this dir.
       - HF_HOME=/app/hf_cache