perf: vectorize tiled fp8 transpose quantize and arm amax via memset

- tiled transpose quantize becomes one 64x32-tile kernel: native pair loads (128B warp reads) with in-kernel scalar fallback at unaligned or ragged rows, so odd widths and misaligned bases no longer route to a separate kernel
- the old 32x32 scalar tiled kernel and its launcher correctness branch are gone; grid sizing simplifies to 1 + total / (vec * threads) since both elementwise loops are grid-stride
- quantize arms the amax buffer with cudaMemsetAsync instead of the zeros() fill kernel, dropping one tensor-op dispatch and kernel launch per call
- byte-exact parity holds over 1404 golden records (13 shapes x 3 dtypes x 3 scales x 2 formats x 3 layouts x aligned/misaligned) and tests/extension passes 65/65
- elementwise quantize kernel left unchanged: 16B-store pairing, __ldcs streaming hints and amax tree reduction all measured neutral at its ~52% DRAM ceiling and were reverted

Benchmark: L20 (sm_89), profiler kernel time with L2 flushed between calls.
- transposed quantize (layout 1): 230 -> 294 GB/s on 2048x1536 (+28%), 245 -> 299 on 2048x1536 weights (+22%); dual-layout (layout 2) 248 -> 329 (+33%) on the same shapes
- DRAM-saturated sizes (~10.6M elements) regress ~5% (404 -> 384 GB/s on 8192x1536), ~0.02% of a training step; accepted for the single-kernel shape after scalar-path and geometry variants both measured the same
- amax init fill kernel 3.0us -> memset 0.9us; quantize call CPU wall 18.5 -> 13.4us on 128x1536

csrc/kernels/fp8/ops.cu

@@ -120,7 +120,11 @@ py::object quantize(torch::Tensor x, torch::Tensor scale, int64_t fmt,
     auto input = x.contiguous();
     auto out_opts = input.options().dtype(
         fmt ? torch::kFloat8_e5m2 : torch::kFloat8_e4m3fn);
-    auto amax = torch::zeros({1}, input.options().dtype(torch::kFloat32));
+    // amax is reduced via atomicMax of non-negative values; a driver memset
+    // arms it cheaper than the zeros() fill kernel (one fewer tensor-op
+    // dispatch + kernel launch on every quantize call).
+    auto amax = torch::empty({1}, input.options().dtype(torch::kFloat32));
+    cudaMemsetAsync(amax.data_ptr(), 0, sizeof(float), stream.stream());
 
     FP8QuantizeParams p;
     p.input_ptr = input.data_ptr();