feat!: providers use unified 'persistence' field

BREAKING CHANGE: Provider config field names changed for semantic clarity

- Rename kvstore → persistence for all providers
- Simple providers: flat persistence with backend reference
- Complex providers (agents): nested persistence.agent_state + persistence.responses
- Files provider: metadata_store → persistence
- Provider configs now clearly express 'how do I persist?' not 'what type of store?'

Example:
  # Before
  config:
    kvstore:
      backend: kvstore
      namespace: faiss

  # After
  config:
    persistence:
      backend: kvstore
      namespace: faiss

  # Agents (nested)
  config:
    persistence:
      agent_state:
        backend: kvstore
        namespace: agents
      responses:
        backend: sqlstore
        namespace: responses

llama_stack/providers/inline/eval/meta_reference/config.py

@@ -14,12 +14,12 @@ from llama_stack.providers.utils.kvstore.config import (
 
 
 class MetaReferenceEvalConfig(BaseModel):
-    kvstore: KVStoreConfig
+    persistence: KVStoreConfig
 
     @classmethod
     def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> dict[str, Any]:
         return {
-            "kvstore": SqliteKVStoreConfig.sample_run_config(
+            "persistence": SqliteKVStoreConfig.sample_run_config(
                 __distro_dir__=__distro_dir__,
                 db_name="meta_reference_eval.db",
             )

llama_stack/providers/inline/eval/meta_reference/eval.py

@@ -54,7 +54,7 @@ class MetaReferenceEvalImpl(
         self.benchmarks = {}
 
     async def initialize(self) -> None:
-        self.kvstore = await kvstore_impl(self.config.kvstore)
+        self.kvstore = await kvstore_impl(self.config.persistence)
         # Load existing benchmarks from kvstore
         start_key = EVAL_TASKS_PREFIX
         end_key = f"{EVAL_TASKS_PREFIX}\xff"