[Evals API][3/n] scoring_functions / scoring meta-reference implementations

[yanxi0830]

---continuation of https://github.com/meta-llama/llama-stack/pull/288

TL;DR

• scoring depends on datasetio & datasets
• Enforce dataset_schema, client enforces dataset schema to follow specified column names/types
• Add clients for datasets / datasetio / scoring
• Basic meta-reference impl for scoring/score, scoring/score_batch

Next PR

• full evals benchmark task with generation
• LLM as judge scorer impl

Tests

unit tests

PROVIDER_ID=test-meta PROVIDER_CONFIG=llama_stack/providers/tests/scoring/provider_config_example.yaml pytest -s llama_stack/providers/tests/scoring/test_scoring.py --tb=short --disable-warnings

PROVIDER_ID=test-meta PROVIDER_CONFIG=llama_stack/providers/tests/datasetio/provider_config_example.yaml pytest -s llama_stack/providers/tests/datasetio/test_datasetio.py --tb=short --disable-warnings

client tests

python -m llama_stack.apis.datasetio.client $LOCALHOST 5000

python -m llama_stack.apis.datasets.client $LOCALHOST 5000

python -m llama_stack.apis.scoring.client $LOCALHOST 5000