[FEATURES] Enhancing Training/Serving Workflow Efficiency

[SecretSun]

What feature you'd like to add:

There is currently a disconnect between submitting training tasks and data pipelines. We would like to support a seamless and efficient end-to-end linkage capability from creating Fluid Cache to triggering preheating to task submission.

Why is this feature needed:

This feature is essential for enhancing the efficiency of algorithm training tasks and aims to shield the user from underlying complexities.