Frame issue in `mjCBody::AccumulateInertia` causing issues with `fusestatic`

### Intro

Hi!

I'm using mujoco to simulate robots. I ran into an issue with loading URDFs containing fixed joints that I think is a bug in adding together inertia.

### My setup

I'm using mujoco built from master on commit e07adc387a09d08d456a8e23336875216c0b5059 on Ubuntu Linux 22.04 x86_64. I've also observed the issue on release 3.3.7. I think the issue will happen on all platforms.

### What's happening? What did you expect?

I noticed the inertias are not added together correctly when `fusestatic` is enabled and the child body is rotated. Translation is handled as expected.

I believe the issue is introduced in this commit: https://github.com/google-deepmind/mujoco/commit/7932b4b202289a98d5bc9b8041d576a091ccddd3

In particular, line 2266 should use `other_iquat` (which includes the pose of the child body) instead of `other->iquat`. 
https://github.com/google-deepmind/mujoco/blob/e07adc387a09d08d456a8e23336875216c0b5059/src/user/user_objects.cc#L2244-L2267

### Steps for reproduction

1. Load the model below, which is a cross and has `fusestatic` set to false.
2. Load in `simulate`. Press `space` to pause, `backspace` to reset the model to the origin, then press `i` to visualize inertias. They look correct.
3. Modify the model to set `fusestatic` to `true`
4. Load in `simulate` and observe the wrong inertia

Here is the model with `fusestatic` set to false:

<img width="476" height="432" alt="Image" src="https://github.com/user-attachments/assets/5f8a24a4-68a3-4ac2-b88b-d0ca175464f1" />

Here is the model with fusestatic set to true. The inertia of the child body (blue) is added to the inertia of the parent body (red) incorrectly. The expected behavior is pizza-box-shaped inertia that goes around the cross.

<img width="592" height="382" alt="Image" src="https://github.com/user-attachments/assets/b8844ad4-ede4-421e-a6bf-02be0b3485cb" />

Next, change line 2266 in `user_objects.cc` to 
```cpp
    {other_iquat[0], other_iquat[1], other_iquat[2], other_iquat[3]}
```
rebuild mujoco, load the model, view inertia and see that the inertia looks correct:

<img width="445" height="450" alt="Image" src="https://github.com/user-attachments/assets/8222ea5b-bd5a-4fdb-b260-3187fe3963c5" />

### Minimal model for reproduction

```XML
<mujoco model="inertia_bug">
  <compiler fusestatic="true" angle="degree"/>

  <visual>
    <rgba inertia="1 1 1 0.2"/>
  </visual>
  <worldbody>
    <body>
      <freejoint/>
      <geom type="capsule" fromto="-1 0 0  1 0 0" size="0.05" rgba="0.8 0.2 0.2 1"/>
      <inertial pos="0 0 0" mass="1" diaginertia="0.01 1 1"/>
      <body pos="0 0 0" euler="0 90 0">
        <geom type="capsule" fromto="-1 0 0  1 0 0" size="0.05" rgba="0.2 0.2 0.8 1"/>
        <inertial pos="0 0 0" mass="1" diaginertia="0.01 1 1"/>
        </body>
    </body>
  </worldbody>
</mujoco>
```


### Code required for reproduction

no code is needed, you can load the model in the `simulate` binary to see the issue.


### Confirmations

- [x] I searched the [latest documentation](https://mujoco.readthedocs.io/en/latest/overview.html) thoroughly before posting.
- [x] I searched previous [Issues](https://github.com/google-deepmind/mujoco/issues) and [Discussions](https://github.com/google-deepmind/mujoco/discussions), I am certain this has not been raised before.

	// body_ipose = body_pose * body_ipose
	double other_ipos[3];
	double other_iquat[4];
	mjuu_copyvec(other_ipos, other->ipos, 3);
	mjuu_copyvec(other_iquat, other->iquat, 4);
	mjuu_frameaccum(other_ipos, other_iquat, other->pos, other->quat);

	// organize data
	double mass[2] = {
	result->mass,
	other->mass
	};
	double inertia[2][3] = {
	{result->inertia[0], result->inertia[1], result->inertia[2]},
	{other->inertia[0], other->inertia[1], other->inertia[2]}
	};
	double ipos[2][3] = {
	{result->ipos[0], result->ipos[1], result->ipos[2]},
	{other_ipos[0], other_ipos[1], other_ipos[2]}
	};
	double iquat[2][4] = {
	{result->iquat[0], result->iquat[1], result->iquat[2], result->iquat[3]},
	{other->iquat[0], other->iquat[1], other->iquat[2], other->iquat[3]}
	};

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Frame issue in `mjCBody::AccumulateInertia` causing issues with `fusestatic` #2982

Intro

My setup

What's happening? What did you expect?

Steps for reproduction

Minimal model for reproduction

Code required for reproduction

Confirmations

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Frame issue in mjCBody::AccumulateInertia causing issues with fusestatic #2982

Description

Intro

My setup

What's happening? What did you expect?

Steps for reproduction

Minimal model for reproduction

Code required for reproduction

Confirmations

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Frame issue in `mjCBody::AccumulateInertia` causing issues with `fusestatic` #2982