Add a simple experimental TorchCompileModel node.

It probably only works on Linux.

For maximum speed on Flux with Nvidia 40 series/ada and newer try using
this node with fp8_e4m3fn and the --fast argument.

nodes.py

@@ -2102,6 +2102,7 @@ def init_builtin_extra_nodes():
         "nodes_hunyuan.py",
         "nodes_flux.py",
         "nodes_lora_extract.py",
+        "nodes_torch_compile.py",
     ]
 
     import_failed = []