[WIP] Add fused rope function for PyTorch

[Copilot]

• Explore repository structure and existing ops patterns
• Study CuTeDSL API from NVIDIA/cutlass examples
• Implement fused RoPE using CuTeDSL in rynn_scale/ops/fused_rope.py
• Export fused_rope from rynn_scale/ops/__init__.py
• Validate implementation with code review and security scan

Original prompt

帮我用cutedsl写一个fused rope的pytorch function，要求：

1. 输入：
   q，shape为(batch_size, seq_len, num_q_heads, head_dim)
   k，shape为(batch_size, seq_len, num_kv_heads, head_dim)
   v，shape为(batch_size, seq_len, num_kv_heads, head_dim)
   cos，shape为(batch_size, seq_len, head_dim)
   sin，shape为(batch_size, seq_len, head_dim)

2. 输出：
   q，k

3. 需要支持inplace操作

创建自 VS Code。

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}