Split safety into (llama-guard, prompt-guard, code-scanner) (#400)

Splits the meta-reference safety implementation into three distinct providers:

- inline::llama-guard
- inline::prompt-guard
- inline::code-scanner

Note that this PR is a backward incompatible change to the llama stack server. I have added deprecation_error field to ProviderSpec -- the server reads it and immediately barfs. This is used to direct the user with a specific message on what action to perform. An automagical "config upgrade" is a bit too much work to implement right now :/

(Note that we will be gradually prefixing all inline providers with inline:: -- I am only doing this for this set of new providers because otherwise existing configuration files will break even more badly.)

llama_stack/providers/remote/inference/bedrock/__init__.py

@@ -3,11 +3,12 @@
 #
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
-from .bedrock import BedrockInferenceAdapter
 from .config import BedrockConfig
 
 
 async def get_adapter_impl(config: BedrockConfig, _deps):
+    from .bedrock import BedrockInferenceAdapter
+
     assert isinstance(config, BedrockConfig), f"Unexpected config type: {type(config)}"
 
     impl = BedrockInferenceAdapter(config)

llama_stack/providers/remote/inference/ollama/ollama.py

@@ -80,6 +80,7 @@ class OllamaInferenceAdapter(Inference, ModelsProtocolPrivate):
                 continue
 
             llama_model = ollama_to_llama[r["model"]]
+            print(f"Found model {llama_model} in Ollama")
             ret.append(
                 Model(
                     identifier=llama_model,