add fixed tendons to aloha pot scene

[thowell]

based on a recommendation from @yuvaltassa, this pr adds fixed tendons to the aloha pot scene.

mjwarp-testspeed mujoco_warp/benchmark/aloha_pot/scene.xml --nconmax=200000 --njmax=128 --replay=lift_pot --measure_solver

main

Rolling out 1000 steps at dt = 0.002...

Summary for 8192 parallel rollouts

Total JIT time: 0.62 s
Total simulation time: 5.65 s
Total steps per second: 1,450,332
Total realtime factor: 2,900.66 x
Total time per step: 689.50 ns
Total converged worlds: 8192 / 8192

solver niter:

mean     std       min  max
---------------------------
2.83956  0.944506    1    9
 2.8739  0.983181    1   10
3.97102   1.69136    1   14
5.48658   1.68678    1   15
 5.7203   1.78359    1   15
5.56193   2.06369    1   16
4.49976   2.10049    1   15
4.59003   1.96566    1   14
4.44851   1.57807    1   15
3.52524   1.35555    1   12

this pr

Summary for 8192 parallel rollouts

Total JIT time: 0.61 s
Total simulation time: 5.68 s
Total steps per second: 1,441,120
Total realtime factor: 2,882.24 x
Total time per step: 693.90 ns
Total converged worlds: 8192 / 8192

solver niter:

mean     std       min  max
---------------------------
2.83318  0.939307    1    9
2.87934  0.985099    1   10
 3.9531   1.68004    1   15
5.51009   1.71513    1   15
5.75248   1.81376    1   15
5.64862   2.09079    1   15
4.54646   2.12754    1   15
4.61883   1.97576    1   15
4.47117   1.56202    1   13
  3.556   1.36036    1   11

[erikfrey]

I like this, but I would pursue this change within Menagerie first, which is tricky as it means testing the new model with a few stakeholders first. This may or may not be worth the effort - we can discuss.

I would not fork the model here away from the official one - the goal of this benchmark is to represent what our users will encounter with MJWarp in the wild. From that perspective, using the model as is in Menagerie is more appropriate.

[erikfrey]

Shouldn't adding tendons mean we can remove the equality constraints?

[thowell]

@yuvaltassa?