Improve Returns section formatting in lu function

Replace bold asterisk formatting with SciPy-style conditional returns
formatting using pipe notation:
- Use "(p, l, u) | (pl, u):" to show alternative return tuples
- Explain conditions in prose followed by "where:"
- List all return values together (p, l, pl, u) for clarity

Also add missing period to parameter description in lu_solve.

This follows SciPy's exact formatting convention for conditional returns,
ensuring proper Sphinx/Napoleon rendering.

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

Author: antonwolfy

dpnp/scipy/linalg/_decomp_lu.py

@@ -95,28 +95,25 @@ def lu(
 
     Returns
     -------
-    **(If ``permute_l`` is ``False``)**
-
-    p : (..., M, M) dpnp.ndarray or (..., M) dpnp.ndarray
-        If `p_indices` is ``False`` (default), the permutation matrix.
-        The permutation matrix always has a real dtype (``float32`` or
-        ``float64``) even when `a` is complex, since it only contains
-        0s and 1s.
-        If `p_indices` is ``True``, a 1-D (or batched) array of row
-        permutation indices such that ``A = L[p] @ U``.
-    l : (..., M, K) dpnp.ndarray
-        Lower triangular or trapezoidal matrix with unit diagonal.
-        ``K = min(M, N)``.
-    u : (..., K, N) dpnp.ndarray
-        Upper triangular or trapezoidal matrix.
-
-    **(If ``permute_l`` is ``True``)**
-
-    pl : (..., M, K) dpnp.ndarray
-        Permuted ``L`` matrix: ``pl = P @ L``.
-        ``K = min(M, N)``.
-    u : (..., K, N) dpnp.ndarray
-        Upper triangular or trapezoidal matrix.
+    (p, l, u) | (pl, u) :
+        The tuple ``(p, l, u)`` is returned if ``permute_l`` is ``False``
+        (default), else the tuple ``(pl, u)`` is returned, where:
+
+        p : (..., M, M) dpnp.ndarray or (..., M) dpnp.ndarray
+            If `p_indices` is ``False`` (default), the permutation matrix.
+            The permutation matrix always has a real dtype (``float32`` or
+            ``float64``) even when `a` is complex, since it only contains
+            0s and 1s.
+            If `p_indices` is ``True``, a 1-D (or batched) array of row
+            permutation indices such that ``A = L[p] @ U``.
+        l : (..., M, K) dpnp.ndarray
+            Lower triangular or trapezoidal matrix with unit diagonal.
+            ``K = min(M, N)``.
+        pl : (..., M, K) dpnp.ndarray
+            Permuted ``L`` matrix: ``pl = P @ L``.
+            ``K = min(M, N)``.
+        u : (..., K, N) dpnp.ndarray
+            Upper triangular or trapezoidal matrix.
 
     Notes
     -----
@@ -273,7 +270,7 @@ def lu_solve(lu_and_piv, b, trans=0, overwrite_b=False, check_finite=True):
     lu, piv : {tuple of dpnp.ndarrays or usm_ndarrays}
         LU factorization of matrix `a` (..., M, M) together with pivot indices.
     b : {(M,), (..., M, K)} {dpnp.ndarray, usm_ndarray}
-        Right-hand side
+        Right-hand side.
     trans : {0, 1, 2} , optional
         Type of system to solve: