perf: native ixformer decode (v1 ≤32K, v2 >32K) + flash_attn_varlen prefill

Replaces all Python PyTorch fallback attention with native ixformer kernels:

Decode path:
- ≤32K: paged_attention_v1 (5D KV layout, x=8) — verified on real BI-V100
- >32K: paged_attention_v2 (5D→4D permute) — verified 65K+ on real BI-V100
- Removes _forward_decode_pytorch Python fallback entirely

Prefill path (profiling):
- _run_sdpa_fallback now uses ixformer.flash_attn_varlen_func
- head_dim=256 verified correct (diff<0.004) and 1.7x faster than PyTorch
- Falls back to Q-tiling pure-math if ixformer unavailable

Also includes: MoE kernel integration, GDN C++ kernels, diagnostic scripts,
xllm upstream layer/kernel references, .dockerignore cleanup.

All changes verified on real BI-V100 hardware (single card).

ex_engine/csrc/ilu_layers_CMakeLists.txt

@@ -0,0 +1,14 @@
+include(cc_library)
+
+cc_library(
+  NAME
+    ilu_layers
+  HDRS
+    attention.h
+    fused_moe.h
+  SRCS
+    attention.cpp
+    fused_moe.cpp
+  DEPS
+    :common_layers
+)