block_sumsq_over_samples
========================

.. py:function:: t3toolbox.backend.fitting.block_sumsq_over_samples(out, n_w, has_order)

   .. code-block:: python

      def block_sumsq_over_samples(
              out:        NDArray,    # scalar-output residual: (order+1)+W+C (has_order) or W+C
              n_w:        int,        # number of leading W axes (unused -- W + C are summed wholesale)
              has_order:  bool,       # True for the derivative kinds (a leading order axis at axis 0)
      ) -> NDArray:                   # (1, n_order) -- per-order sum of squares (no mode axis: apply/entries)


   Per-``(mode, order)`` sum-of-squares for the scalar-output apply/entries kinds -> a 2-D
   ``(1, n_order)`` matrix (no mode axis -- they contract every mode). Keeps only a leading order axis,
   sums the rest (``W`` + ``C``); **UNWEIGHTED** (raw ``‖r_·j‖²``).