docs: Phase T13 — activation recompute

Design doc for per-block gradient checkpointing (KI-3): the no-tape forward +
recompute-on-backward design, the `checkpoint` primitive, per-block wrapping,
the exactness/correctness argument (same kernels + inputs → identical grads),
composition with bf16+DDP+batched, and the verification plan (on-vs-off grad
gate + memory/throughput before→after, dim1024-fits). Bench table left as TBD
to fill after the dash5 run.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

crates/xtrain-model/tests/recompute.rs

@@ -48,8 +48,13 @@ fn build(cfg: Config, device: Device, dtype: DType, recompute: bool) -> TinyTran
     m.with_compute_dtype(dtype).with_recompute(recompute)
 }
 
+/// Upcast to fp32 then read to host — logits are bf16 in bf16 mode (grads are
+/// always fp32 master, but this is uniform and harmless for fp32 tensors).
 fn host(t: &xtrain_tensor::Tensor) -> Vec<f32> {
-    t.to_device(Device::Cpu).as_slice::<f32>().to_vec()
+    t.to_dtype(DType::F32)
+        .to_device(Device::Cpu)
+        .as_slice::<f32>()
+        .to_vec()
 }
 
 fn run(dtype: DType, logit_tol: f32, grad_tol: f32) {