Add LLaMA model test (#4070)

Summary:
Pull Request resolved: #4070

As titled.

Required changes to make it work:
- replace `aten::full_like` with `aten::full`
- remove `aten::logical_neg` when used as an input to `aten::where`
- replace `-inf` value in `aten::full`

Reviewed By: JacobSzwejbka, cccclai

Differential Revision: D58974035

fbshipit-source-id: c9f8f9f83c4c9a92f84995290b08a74b672d959e

examples/models/llama2/TARGETS

@@ -15,6 +15,7 @@ runtime.python_library(
     base_module = "executorch.examples.models.llama2",
     visibility = [
         "//executorch/...",
+        "@EXECUTORCH_CLIENTS",
     ],
     deps = [
         "//caffe2:torch",