[DOC] Add doc for memory views with custom numpy dtype (GH-2813)

Closes https://github.com/cython/cython/issues/2760

docs/examples/userguide/memoryviews/custom_dtype.pyx

+import numpy as np
+
+CUSTOM_DTYPE = np.dtype([
+    ('x', np.uint8),
+    ('y', np.float32),
+])
+
+a = np.zeros(100, dtype=CUSTOM_DTYPE)
+
+cdef packed struct custom_dtype_struct:
+    # The struct needs to be packed since by default numpy dtypes aren't
+    # aligned
+    unsigned char x
+    float y
+
+def sum(custom_dtype_struct [:] a):
+
+    cdef:
+        unsigned char sum_x = 0
+        float sum_y = 0.
+
+    for i in range(a.shape[0]):
+        sum_x += a[i].x
+        sum_y += a[i].y
+
+    return sum_x, sum_y

docs/src/userguide/memoryviews.rst

@@ -78,6 +78,12 @@ three dimensional buffer into a function that requires a two
 dimensional buffer will raise a ``ValueError``.
 
 
+To use a memory view on a numpy array with a custom dtype, you'll need to
+declare an equivalent packed struct that mimics the dtype:
+
+.. literalinclude:: ../../examples/userguide/memoryviews/custom_dtype.pyx
+
+
 Indexing
 --------