ci(mlx): add LoRA + merged_16bit + GGUF export round-trip checks

After the 7-step LoRA training run finishes and the in-memory
inference assertion passes, the smoke test now exports the trained
model in three formats, drops the in-memory model + trainer to
reclaim memory, and reloads each export from disk to re-run the
"<<HELLO!!>> My name is " inference assertion. Each reload is
expected to still complete with "Unsloth" -- catching round-trip
regressions where the saved weights silently corrupt or fail to
load.

Formats exercised:

- LoRA adapter via model.save_pretrained_merged(save_method="lora").
  Reloaded with FastMLXModel.from_pretrained on the adapter dir;
  the loader auto-detects adapter_config.json and pulls down the
  base model.

- Merged 16-bit via model.save_pretrained_merged(save_method=
  "merged_16bit"). Fuses LoRA into the base, dequantizes to fp16,
  saves an HF-compatible safetensors directory. Reload via
  FastMLXModel.from_pretrained on the saved dir.

- GGUF via model.save_pretrained_gguf(quantization_method=
  "not_quantized"). Builds llama.cpp via cmake on the runner with
  GGML_METAL=ON (only the llama-cli, llama-quantize, and
  llama-gguf-split targets), then runs the produced bf16 GGUF
  through llama-cli with a fixed seed and asserts "Unsloth" in
  stdout. GGUF infra failures (cmake / build / convert) are
  surfaced as RuntimeError so we notice -- if Mac CI starts hitting
  build flakes the assertion can be softened.

Workflow timeout bumped 15 -> 25 min to budget for the llama.cpp
cmake build (~5-7 min on the macos-14 standard runner).

.github/workflows/mlx-ci.yml

@@ -84,7 +84,10 @@ jobs:
   dispatch:
     name: dispatch
     runs-on: macos-14
-    timeout-minutes: 15
+    # 25 min: dispatch + spoofed matrix + 7-step real LoRA training is
+    # under 2 min; GGUF export builds llama.cpp via cmake on Apple
+    # Silicon (~5-7 min), so we budget headroom.
+    timeout-minutes: 25
     steps:
       - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5  # v4.3.1