[MLX] Add MLX support to the Qwen3.5 C++ runner #12381

Job	Run time
Get changed files / get-changed-files	3s
CI run decision / decide	32s
test-metal-qwen35-moe-tiny / macos-job	18m 25s
test-metal-backend-modules / macos-job	18m 25s
test-executorch-metal-build / macos-job	18m 25s
export-model-metal-artifact (mistralai, Voxtral-Mini-4B-Realtime-2602, quantized-int4-metal) / macos-job	18m 26s
export-model-metal-artifact (nvidia, parakeet-tdt, quantized-int4-metal) / macos-job	18m 26s
export-model-metal-artifact (openai, whisper-small, non-quantized) / macos-job	18m 26s
export-model-metal-artifact (mistralai, Voxtral-Mini-3B-2507, quantized-int4-metal) / macos-job	18m 26s
export-model-metal-artifact (openai, whisper-large-v3-turbo, non-quantized) / macos-job	18m 26s
export-model-metal-artifact (nvidia, parakeet-tdt, non-quantized) / macos-job	18m 26s
export-model-metal-artifact (openai, whisper-small, quantized-int4-metal) / macos-job	18m 26s
export-model-metal-artifact (openai, whisper-large-v3-turbo, quantized-int4-metal) / macos-job	18m 26s
export-model-metal-artifact (mistralai, Voxtral-Mini-3B-2507, non-quantized) / macos-job	18m 26s
test-model-metal-e2e	-1s
	3h 41m 43s

Provide feedback