remote rust_local_mesh

This diff removes the old way we created tensor engines via rust local mesh. We were no longer testing this in test_remote_functions, etc. but still have a bunch of old examples that depended on it.

The reason to do this now is to lighten what we actually use in the monarch_tensor_worker crate. We need to sever its direct dependence on torch and getting rid of its dead code makes it clearer what we have to do.

Pull Request resolved: #1947
ghstack-source-id: 324737962

Differential Revision: [D87496994](https://our.internmc.facebook.com/intern/diff/D87496994/)

Author: zdevito

.github/workflows/test-gpu-rust.yml

@@ -66,7 +66,6 @@ jobs:
         timeout 12m cargo nextest run --workspace --profile ci \
           --exclude monarch_messages \
           --exclude monarch_tensor_worker \
-          --exclude monarch_simulator_lib \
           --exclude torch-sys \
           --exclude torch-sys-cuda
         # Copy the test results to the expected location

.gitignore

@@ -10,7 +10,6 @@ python/monarch.egg-info/*
 build/*
 dist/*
 monarch.egg-info/*
-python/monarch/monarch_controller
 
 .ipynb_checkpoints
 .monarch

Cargo.toml

@@ -2,7 +2,6 @@
 resolver = "2"
 members = [
     "build_utils",
-    "controller",
     "cuda-sys",
     "erased_lifetime",
     "hyper",
@@ -18,7 +17,6 @@ members = [
     "monarch_messages",
     "monarch_perfetto_trace",
     "monarch_rdma",
-    "monarch_simulator",
     "monarch_tensor_worker",
     "monarch_types",
     "nccl-sys",