[SPARK-21976][DOC] Fix wrong documentation for Mean Absolute Error.

## What changes were proposed in this pull request?

Fixed wrong documentation for Mean Absolute Error.

Even though the code is correct for the MAE:

```scala
Since("1.2.0")
  def meanAbsoluteError: Double = {
    summary.normL1(1) / summary.count
  }
```
In the documentation the division by N is missing.

## How was this patch tested?

All of spark tests were run.

Please review http://spark.apache.org/contributing.html before opening a pull request.

Author: FavioVazquez <favio.vazquezp@gmail.com>
Author: faviovazquez <favio.vazquezp@gmail.com>
Author: Favio André Vázquez <favio.vazquezp@gmail.com>

Closes #19190 from FavioVazquez/mae-fix.

(cherry picked from commit e2ac2f1)
Signed-off-by: Sean Owen <sowen@cloudera.com>

docs/mllib-evaluation-metrics.md

@@ -549,7 +549,7 @@ variable from a number of independent variables.
     </tr>
     <tr>
       <td>Mean Absolute Error (MAE)</td>
-      <td>$MAE=\sum_{i=0}^{N-1} \left|\mathbf{y}_i - \hat{\mathbf{y}}_i\right|$</td>
+      <td>$MAE=\frac{1}{N}\sum_{i=0}^{N-1} \left|\mathbf{y}_i - \hat{\mathbf{y}}_i\right|$</td>
     </tr>
     <tr>
       <td>Coefficient of Determination $(R^2)$</td>