fix(ci): PTX cache invalidation includes build.rs + all kernel dirs

The script only hashed .cu/.cuh in crates/ml/. Missed:
- build.rs changes (removing --use_fast_math → stale cubins cached)
- crates/ml-dqn/src/*.cu (replay buffer, seg tree kernels)
- crates/ml-core/src/cuda_autograd/*.cu
- crates/ml-ppo/src/cuda_nn/*.cu

Now hashes ALL kernel sources + ALL build.rs files. Any change to
nvcc flags or kernel source invalidates the entire PTX cache.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

scripts/ptx-cache-invalidate.sh

@@ -16,13 +16,17 @@
 
 set -euo pipefail
 
-KERNEL_DIR="crates/ml/src/cuda_pipeline"
+KERNEL_DIRS="crates/ml/src/cuda_pipeline crates/ml-dqn/src crates/ml-core/src/cuda_autograd crates/ml-ppo/src/cuda_nn"
 CACHE_DIR="${1:-${CARGO_TARGET_DIR:+${CARGO_TARGET_DIR}/.ptx_cache}}"
 CACHE_DIR="${CACHE_DIR:-/tmp/.ptx_cache}"
 
-# Compute content hash of all CUDA source files
-HASH=$(find "$KERNEL_DIR" -type f \( -name '*.cu' -o -name '*.cuh' \) \
-       -exec sha256sum {} + 2>/dev/null | sort | sha256sum | cut -d' ' -f1)
+# Compute content hash of CUDA sources AND build.rs (nvcc flags).
+# build.rs changes (e.g. removing --use_fast_math) produce different cubins
+# even when .cu source is unchanged — must invalidate cache.
+HASH=$( {
+    find $KERNEL_DIRS -type f \( -name '*.cu' -o -name '*.cuh' \) -exec sha256sum {} + 2>/dev/null
+    find crates/ml crates/ml-dqn crates/ml-core crates/ml-ppo -maxdepth 1 -name 'build.rs' -exec sha256sum {} + 2>/dev/null
+} | sort | sha256sum | cut -d' ' -f1)
 
 STAMP_FILE="${CACHE_DIR}/.kernel_hash"