refactor: remove bf16 gemm and swiglu kernels and rebuild csrc benchmarks

- delete csrc/kernels/gemm.cu and swiglu.cu and drop their CMake and setup.py registration
- remove the ops wrappers plus backend/linear.py and backend/swiglu.py so Linear and MLP call F.linear directly
- drop the four gemm and swiglu kernel test files and prune the stale cuda_kernels.md sections
- add csrc/bench benchmarks for the remaining kernels: attention decode prefill paged decode paged prefill versus single-launch SDPA references, rotary versus the torch fallback, fp8 quantize and mm_fp8 versus torch baselines
- attention, rotary_emb, and fp8_ops kernels are unchanged

astrai/model/components/linear.py

@@ -1,9 +1,8 @@
 import torch
 import torch.nn as nn
+import torch.nn.functional as F
 from torch import Tensor
 
-from astrai.extension.backend.linear import linear
-
 
 class Linear(nn.Module):
     def __init__(
@@ -22,4 +21,4 @@ class Linear(nn.Module):
             nn.init.uniform_(self.bias, -bound, bound)
 
     def forward(self, x: Tensor) -> Tensor:
-        return linear(x, self.weight, self.bias)
+        return F.linear(x, self.weight, self.bias)