[e2e_template] Update llmforge version for e2e LLM template (#372)

This makes the Ray Data + Ray Train dashboard available for the
template.

I tested it and made sure the model still gets high accuracy and can be
served.

templates/e2e-llm-workflows/deploy/jobs/ft.yaml

@@ -1,6 +1,6 @@
 name: e2e-llm-workflows
 entrypoint: llmforge anyscale finetune configs/training/lora/llama-3-8b.yaml
-image_uri: localhost:5555/anyscale/llm-forge:0.5.4
+image_uri: localhost:5555/anyscale/llm-forge:0.5.7
 requirements: []
 max_retries: 1
 excludes: ["assets"]