Torch compile fusion backend prototype

[bnellnm]

pulls in parts of vllm-project#3014

for now, only the forward method on LlamaMLP is tagged with the new backend. sample code to run (derived from examples/offline_inference.py):

from vllm import LLM, SamplingParams

# to turn up the logging level
import logging
import vllm
vllm.logger._default_handler.setLevel(logging.DEBUG)

# Sample prompts.
prompts = [
    "Hello, my name is",
    "The president of the United States is",
    "The capital of France is",
    "The future of AI is",
]
# Create a sampling params object.
sampling_params = SamplingParams(temperature=0.8, top_p=0.95)

# Create an LLM.
llm = LLM(model="meta-llama/Llama-2-7b-chat-hf")
# Generate texts from the prompts. The output is a list of RequestOutput objects
# that contain the prompt, generated text, and other information.
outputs = llm.generate(prompts, sampling_params)
# Print the outputs.
for output in outputs:
    prompt = output.prompt
    generated_text = output.outputs[0].text
    print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")