Initial project scaffold

tasks/04_online_softmax/cuda_skeleton.cu

@@ -0,0 +1,7 @@
+// Workbook-local CUDA sketch for online softmax.
+//
+// TODO(student):
+// 1. Choose how one block owns one row or row tile.
+// 2. Keep running_max and running_sum across column tiles.
+// 3. Update the recurrence carefully for numerical stability.
+// 4. Normalize the final row.