autodiff: finite-diff gradient-check harness

New xtrain-autodiff crate with a reusable central finite-difference
gradient check: grad_check(x, shape, f, analytic_grad, cfg) compares an
analytic gradient against (f(x+ε)-f(x-ε))/2ε per element with a relative
tolerance. Host-only (no CUDA): the loss closure owns any GPU work, so
T4's per-op backward checks can reuse it directly. Includes host unit
tests (sum(x²) grad 2x passes; a wrong grad is rejected).

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

Cargo.lock

@@ -88,6 +88,13 @@ version = "1.0.24"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "e6e4313cd5fcd3dad5cafa179702e2b244f760991f45397d14d4ebf38247da75"
 
+[[package]]
+name = "xtrain-autodiff"
+version = "0.1.0"
+dependencies = [
+ "xtrain-tensor",
+]
+
 [[package]]
 name = "xtrain-cuda"
 version = "0.1.0"
@@ -101,6 +108,7 @@ version = "0.1.0"
 dependencies = [
  "half",
  "smallvec",
+ "xtrain-autodiff",
  "xtrain-cuda",
 ]