video/image: honor explicit Speed=off for companions + trim, probe explicit TE kernels, bench fidelity

Address the Codex review round on the video/quant work:

- Companion auto-quant now honors an explicit Speed=off. Both loaders already pin the DiT dense
  under an explicit off (bit-exact reference), but the unset text-encoder / VAE quant still promoted
  to auto and silently fp8/int8'd the companions, breaking the bit-exact request. An UNSET speed
  still auto-quantises; an explicit companion scheme still forces it.
- The HunyuanVideo joint-attention trim is a speed lever (it swaps to the fused SDPA kernel), so gate
  it on a non-off speed tier exactly like the adjacent attention-backend selection -- the off path
  keeps the stock dense-mask attention.
- Explicit torchao text-encoder modes (int8 / fp8_dynamic / nvfp4) now run the same kernel smoke
  test the auto ladder uses. They could clear the capability gate yet fail the real GEMM on a build
  where quantize_ wraps the encoder but the kernel is broken; the caster's try/except only covers the
  cast, not the first forward, so the load would report engaged then crash at generation. Now it
  falls back to dense. Layerwise fp8 has no torchao GEMM, so the probe is a no-op for it.
- The trim pre-hook's fallback restores the caller's original kwargs (it may have emptied the image
  stream / trimmed a text stream before failing), so the stock dense-mask path runs on exactly what
  it expects, matching the empty-prompt guard.
- video_speedmem_bench mirrors the loader: installs the Hunyuan trim before the backend set (gated on
  an active tier) and skips the auto int8 quant when it is the fp8-denied memory fallback and dense
  fits resident, so the shipped/auto rows measure what the loader actually runs.

Tests: TE explicit-mode kernel probe (+ layerwise-fp8 bypass), trim mid-trim restore, and loader-level
speed=off companion suppression + trim skip for both backends. 262 backend tests pass; ruff clean.

studio/backend/tests/test_diffusion_backend.py

@@ -2273,6 +2273,31 @@ def test_speed_off_load_suppresses_auto_dtype_quant(fake_runtime, tmp_path, monk
     assert _FakeTransformer.last["path"]  # GGUF from_single_file was used, not a dense build
 
 
+def test_speed_off_load_suppresses_auto_companion_quant(fake_runtime, tmp_path, monkeypatch):
+    # Mirror the DiT suppression for the companions: an explicit Speed="off" load with TE/VAE left at
+    # auto must keep them dense (mode "off"), not promote them to auto-quant and silently fp8/int8 the
+    # text encoder + VAE, which would break the bit-exact request. Unset speed still auto-quantises.
+    from core.inference import diffusion as dmod
+
+    te_modes: list = []
+    vae_modes: list = []
+    monkeypatch.setattr(
+        dmod, "quantize_text_encoders", lambda pipe, target, *, mode, **kw: te_modes.append(mode)
+    )
+    monkeypatch.setattr(
+        dmod, "quantize_vae", lambda pipe, target, *, mode, **kw: vae_modes.append(mode)
+    )
+    (tmp_path / "m.gguf").write_bytes(b"x")
+    backend = DiffusionBackend()
+    backend.load_pipeline(
+        str(tmp_path), gguf_filename = "m.gguf", family_override = "z-image", speed_mode = "off"
+    )
+    assert te_modes == ["off"] and vae_modes == ["off"]  # dense, not auto
+    backend.unload()
+    backend.load_pipeline(str(tmp_path), gguf_filename = "m.gguf", family_override = "z-image")
+    assert te_modes[-1] == "auto" and vae_modes[-1] == "auto"  # promoted when speed is not off
+
+
 def test_transformer_quant_dense_path_engaged(fake_runtime, tmp_path, monkeypatch):
     # transformer_quant + a CUDA resident plan -> load the DENSE transformer from the
     # base repo, place it on the device, quantise it, and report the engaged scheme.