add batch support for contractions

[mganahl]

We lost this ability when removing experiments and BatchTensorNetwork. I think we should add it back.

[chaserileyroberts]

Is this something beyond what tf.vectorized_map and jax.vmap can do?

[mganahl]

actually, it seems no. I'll close this then

[chaserileyroberts]

It would still be useful in a numpy backend.

[mganahl]

right, I see.

[jemisjoky]

It would still be useful in a numpy backend.

Supporting batch contractions with a pytorch backend would be really useful too!