Update cummax documentation (#3323)

Right now, the cummax() documentation uses [1, 2, 3] as its example series. This series isn't the best short example because its cummax is also [1, 2, 3]. It's not immediately clear why passing that as input gets you identical output. The reason the output is identical to the input is that [1, 2, 3] happens to be a monotonically increasing sequence.

The slight change I'm introducing here is an example Series that's *not* monotonically increasing. I think this choice might help users better understand what cummax actually does.

py-polars/polars/internals/series.py

@@ -1109,14 +1109,14 @@ def cummax(self, reverse: bool = False) -> "Series":
 
         Examples
         --------
-        >>> s = pl.Series("a", [1, 2, 3])
+        >>> s = pl.Series("a", [3, 5, 1])
         >>> s.cummax()
         shape: (3,)
         Series: 'a' [i64]
         [
-            1
-            2
             3
+            5
+            5
         ]
 
         """