feat: audio surfaces (TTS + transcription), imagegen.Editor, llamaswap health probe

- New leaf package `audio` (ADR-0017): SpeechModel/SpeechProvider and
  TranscriptionModel/TranscriptionProvider with imagegen conventions
  (zero value = backend default, functional options + Apply, bytes
  in/out, never URLs). Root re-exports added.
- imagegen.Editor (ADR-0018): optional image-to-image interface —
  EditRequest carries the generation knobs plus Init image and
  denoising Strength; separate interface so existing Models keep
  compiling.
- provider/llamaswap implements all of it: POST /v1/audio/speech (JSON,
  raw-audio response, MIME from Content-Type with format fallback),
  POST /v1/audio/transcriptions (multipart, response_format=json),
  ListVoices (GET /v1/audio/voices?model=, tolerant of string-list and
  object-list shapes), POST /sdapi/v1/img2img (txt2img wire +
  init_images/denoising_strength, shared image decode), and Health(ctx)
  (GET /health) — a cheap liveness probe for often-offline hosts.
- Hermetic httptest coverage for every new wire shape and validation
  path; README sections + support-matrix footnote updated in the same
  commit (also corrects the stale /v1/images/generations claim — the
  image path has been SDAPI since the seed fix).

First consumer: mort's llamaswap media tool cluster (status / image /
TTS / STT agent tools against the netherstorm host).

Co-Authored-By: Claude Fable 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01AXQxVhXBw8PwFAtsVrXSmj

provider/llamaswap/health.go

@@ -0,0 +1,45 @@
+package llamaswap
+
+import (
+	"context"
+	"fmt"
+	"io"
+	"net/http"
+)
+
+// Health reports whether the llama-swap instance is reachable (GET
+// {base}/health, which answers "OK" without touching any model). A nil error
+// means the proxy itself is up — it says nothing about how long a subsequent
+// request will take (a cold model swap can still block for minutes). Bound
+// the probe with a short context deadline; the client has no timeout by
+// design.
+func (p *Provider) Health(ctx context.Context) error {
+	if p.baseURL == "" {
+		return fmt.Errorf("llama-swap provider %q: no base URL configured (set one via WithBaseURL or an LLM_* env DSN)", p.name)
+	}
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, p.baseURL+"/health", nil)
+	if err != nil {
+		return fmt.Errorf("llama-swap: build request: %w", err)
+	}
+	if p.token != "" {
+		req.Header.Set("Authorization", "Bearer "+p.token)
+	}
+	resp, err := p.client.Do(req)
+	if err != nil {
+		return fmt.Errorf("llama-swap: health probe: %w", err)
+	}
+	defer resp.Body.Close()
+	_, _ = io.Copy(io.Discard, io.LimitReader(resp.Body, 1<<10))
+	if resp.StatusCode/100 != 2 {
+		return &apiHealthError{status: resp.StatusCode}
+	}
+	return nil
+}
+
+// apiHealthError distinguishes "reachable but unhealthy" from transport
+// failure without pulling in the llm error taxonomy for a probe.
+type apiHealthError struct{ status int }
+
+func (e *apiHealthError) Error() string {
+	return fmt.Sprintf("llama-swap: health endpoint returned status %d", e.status)
+}