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/registry/memory.py

@@ -38,11 +38,11 @@ def available_providers() -> List[ProviderSpec]:
             pip_packages=EMBEDDING_DEPS + ["faiss-cpu"],
             module="llama_stack.providers.inline.memory.faiss",
             config_class="llama_stack.providers.inline.memory.faiss.FaissImplConfig",
-            deprecation_warning="Please use the `faiss` provider instead.",
+            deprecation_warning="Please use the `inline::faiss` provider instead.",
         ),
         InlineProviderSpec(
             api=Api.memory,
-            provider_type="faiss",
+            provider_type="inline::faiss",
             pip_packages=EMBEDDING_DEPS + ["faiss-cpu"],
             module="llama_stack.providers.inline.memory.faiss",
             config_class="llama_stack.providers.inline.memory.faiss.FaissImplConfig",