Added python bindings for iiwa generated grid.cuh file

[kawotwi]

Changes

This PR adds Python bindings for CUDA-accelerated rigid body dynamics algorithms using pybind11. The goal is to expose key GRiD functionality—like inverse dynamics—to Python for easier integration, testing, and scripting.

---

Highlights

• Exposes CUDA logic to Python via pybind11
• Adds GRidDataFloat Python class in pybind logic
• Implements .load_joint_info() and .inverse_dynamics()as well as others included in grid.cuh within this PR.
• installs via pip install -e .
• Integrates CMake with setuptools (setup.py)
• Separates device-only code to avoid host compilation errors

---

How to Use

Install:

cd bindings
pip install -e .

Basic usage:

import gridCuda
import numpy as np

q = np.zeros(gridCuda.NUM_JOINTS, dtype=np.float32)
qd = np.zeros(gridCuda.NUM_JOINTS, dtype=np.float32)
u = np.zeros(gridCuda.NUM_JOINTS, dtype=np.float32)

g = gridCuda.GRidDataFloat()
g.load_joint_info(q, qd, u) # customizable can do random as well. 
c = g.inverse_dynamics()

---

Notes

• Currently supports fixed-base robots only for now based on logic that checks joint length sizes but this can be a q uick change.
• CUDA resources (streams, device memory) are initialized and released properly.
• Resolved issues with host-side compilation of device-only symbols like blockDim and __syncthreads.

---

Next Steps

• Add unit tests for Python bindings
• Document edge cases and expected input formats
• Add performance benchmarks and profiling tools

---

Questions / Discussion

If anyone has thoughts on how to extend this to batched dynamics calls or better handle stream management, please share.

---

📁 Directory Changes

GRiD/
├── bindings/
│   ├── grid_cuda.cpp        # pybind11 wrapper
│   ├── setup.py             # builds shared object via CMake
│   ├── CMakeLists.txt
│   └── ...
├── include/
│   └── grid.cuh             # CUDA dynamics logic

---