Warp + Madrona Support in Playground

Hi @btaba @StafaH ! I'm wondering if Madrona batch rendering is supported yet (or will be supported soon) with MuJoCo Warp?

I switched from the JAX backend to Warp for experiments with MuJoCo Playground, but i'd like to train policies with vision inputs. If Warp+Madrona isn't available yet for vision-based RL, is it still recommended we use JAX backend until the next announcement?

I also noticed that the commits [in this closed PR](https://github.com/google-deepmind/mujoco_warp/pull/735) weren't merged into `main` for `mujoco_warp` 👀 Is there a reason why? (e.g. missing functionalities) I saw activity here https://github.com/StafaH/mujoco_warp/tree/render_context in one of @StafaH 's branches that may be related.