[ExecuTorch][WebGPU] SDPA test suite: replay + dynamic input_pos + in-graph KV cache

Pull Request resolved: pytorch#20087

Adds the WebGPU SDPA test coverage as its own diff, stacked on the SDPA op (which already carries the dynamic-`input_pos` consumption) and the SymInt mechanism below it: multi-step prefill->mt->decode replay, runtime-dynamic `input_pos` (autoregressive decode), and an in-graph mutable KV cache, each compared against a torch `F.scaled_dot_product_attention` golden.

- `test/ops/sdpa/test_sdpa.py`: `ReplaySeq`/`REPLAY_SEQS` + per-step replay export/golden; `DynamicSdpaModule` + `export_dynamic_decode` (one `.pte`, `input_pos` supplied at runtime as a SymInt); `DecodeCacheModule` + `export_incache_decode` (KV cache as `register_buffer` mutable buffers, so the cache persists in-graph and forward() feeds only the new token + `input_pos`).
- `test/test_webgpu_native.cpp`: `test_sdpa_replay`, `test_sdpa_dynamic_decode` (+ negative control: a pinned `input_pos` diverges), `test_sdpa_incache_decode` (+ static control: a fresh Module per step diverges, proving in-graph accumulation is real), `test_symint_roundtrip`, `test_resize_hook`; shared per-element tolerance `sdpa_within_tol` (abs 1e-4 OR rel 1e-3).
- `test/test_build_webgpu.sh`: export the replay / dynamic / in-graph-cache models for the native test.
Authored with assistance from Claude.

ghstack-source-id: 393014582
@exported-using-ghexport

Differential Revision: [D107595144](https://our.internmc.facebook.com/intern/diff/D107595144/)

Author: JCNTH

backends/webgpu/scripts/test_webgpu_native_ci.sh

@@ -75,6 +75,21 @@ export_update_cache_replay('${UPDATE_CACHE_DIR}')
 export_update_cache_negative('${UPDATE_CACHE_DIR}')
 " || { echo "WARN: update_cache export failed; skipping update_cache native test"; UPDATE_CACHE_OK=0; }
 
+# Non-fatal: a failed sdpa export makes the required 4k/8k configs hard-fail in
+# webgpu_native_test below (precise per-config error), so don't exit/mask here.
+$PYTHON_EXECUTABLE -c "
+from executorch.backends.webgpu.test.ops.sdpa.test_sdpa import (
+    export_all_sdpa_models,
+    export_replay_sequences,
+    export_dynamic_decode,
+    export_incache_decode,
+)
+export_all_sdpa_models('/tmp')
+export_replay_sequences('/tmp')
+export_dynamic_decode('/tmp')
+export_incache_decode('/tmp')
+" || echo "WARN: sdpa export failed; required 4k/8k configs will FAIL in webgpu_native_test"
+
 # ── Configure (Dawn-only: no -DWEBGPU_IMPL; Dawn is the sole backend) ─────────
 echo "=== Configure WebGPU native tests on Dawn ==="
 rm -rf "${BUILD_DIR}"