fix: resolve type hint issues and import dependencies (#1176)

# What does this PR do?

- Fixed type hinting and missing imports across multiple modules.
- Improved compatibility by using `TYPE_CHECKING` for conditional
imports.
- Updated `pyproject.toml` to enforce stricter linting.

Signed-off-by: Sébastien Han <seb@redhat.com>

Signed-off-by: Sébastien Han <seb@redhat.com>

llama_stack/providers/inline/inference/meta_reference/parallel_utils.py

@@ -231,7 +231,7 @@ def worker_process_entrypoint(
     while True:
         try:
             task = req_gen.send(result)
-            if isinstance(task, str) and task == _END_SENTINEL:
+            if isinstance(task, str) and task == EndSentinel():
                 break
 
             assert isinstance(task, TaskRequest)

llama_stack/providers/inline/inference/meta_reference/quantization/loader.py

@@ -12,6 +12,7 @@ import os
 from typing import Any, Dict, List, Optional
 
 import torch
+from fairscale.nn.model_parallel.initialize import get_model_parallel_rank
 from fairscale.nn.model_parallel.layers import ColumnParallelLinear, RowParallelLinear
 from fairscale.nn.model_parallel.mappings import reduce_from_model_parallel_region
 from llama_models.llama3.api.args import ModelArgs

llama_stack/providers/inline/tool_runtime/code_interpreter/matplotlib_custom_backend.py

@@ -73,7 +73,10 @@ def show():
         image_data.append({"image_base64": image_base64})
         buf.close()
 
-    req_con, resp_con = _open_connections()
+    # The _open_connections method is dynamically made available to
+    # the interpreter by bundling code from "code_env_prefix.py" -- by literally prefixing it -- and
+    # then "eval"ing it within a sandboxed interpreter.
+    req_con, resp_con = _open_connections()  # noqa: F821
 
     _json_dump = _json.dumps(
         {