Process impl flag for training

[kywch]

Hello, thank you for developing MuJoCo Playground!

I wanted to test Warp and found that the impl flag was not being used to create the training environment. After making these changes, I could use the JAX and Warp backends for training with the impl flag.

Thank you!

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[kywch]

With the warp backend, I'm getting the following messages, as you mentioned, when training the default LeapCubeReorient env. Increasing nxmax in the leap_hand env files fixed this, but I did not include it in this PR.

nefc overflow - please increase njmax to 148
nefc overflow - please increase njmax to 129
nefc overflow - please increase njmax to 140
nefc overflow - please increase njmax to 132
nefc overflow - please increase njmax to 132
nefc overflow - please increase njmax to 132
nefc overflow - please increase njmax to 132
nefc overflow - please increase njmax to 132
nefc overflow - please increase njmax to 132
nefc overflow - please increase njmax to 136
nefc overflow - please increase njmax to 136