[torchtune integration] post training + eval (#670)

## What does this PR do?

- Add related Apis in experimental-post-training template to enable eval
on the finetuned checkpoint in the template
- A small bug fix on meta reference eval
- A small error handle improvement on post training 


## Test Plan
From client side issued an E2E post training request
https://github.com/meta-llama/llama-stack-client-python/pull/70 and get
eval results successfully

<img width="1315" alt="Screenshot 2024-12-20 at 12 06 59 PM"
src="https://github.com/user-attachments/assets/a09bd524-59ae-490c-908f-2e36ccf27c0a"
/>

llama_stack/templates/experimental-post-training/build.yaml

@@ -4,10 +4,22 @@ distribution_spec:
   description: Experimental template for post training
   docker_image: null
   providers:
+    inference:
+    - inline::meta-reference
+    eval:
+    - inline::meta-reference
+    scoring:
+    - inline::basic
     post_training:
     - inline::torchtune
     datasetio:
     - remote::huggingface
     telemetry:
     - inline::meta-reference
+    agents:
+    - inline::meta-reference
+    safety:
+    - inline::llama-guard
+    memory:
+    - inline::faiss
 image_type: conda