Feat/value response

[akshay18iitg]

What this does

Adds response to value function so it can be co-trained on robotic as well as VQA dataset.

Important as mentioned in pi*0.6 paper that co-training value function on VQA dataset helps to avoid overfitting.

How it was tested

Run training on local desktop on mixture of libero and pixmo dataset.

Examples:
opentau-train --config_path=configs/examples/value_config.json

How to checkout & try? (for the reviewer)

opentau-train --config_path=configs/examples/value_config.json

Checklist

• I have added Google-style docstrings to important functions and ensured function parameters are typed.
• My PR includes policy-related changes.
  • If the above is checked: I have run the GPU pytests (pytest -m "gpu") and regression tests.

Note: Before submitting this PR, please read the contributor guideline.