feat: replace seaborn with matplotlib for box_plot (#863)

Closes #805
Closes #849 

### Summary of Changes

This pull request applies the changes of replacing the last import of
seaborn with matplotlib, so that seaborn will be removed as dependency
with this pull request.

We also replaced the snapshot images of the box_plot test, to ensure
that the test run successfully.

---------

Co-authored-by: thelenf0 <[email protected]>
Co-authored-by: megalinter-bot <[email protected]>

pyproject.toml

@@ -19,7 +19,6 @@ matplotlib = "^3.6.3"
 pillow = ">=9.5,<11.0"
 polars = {extras = ["numpy", "pyarrow"], version = "^0.20.25"}
 scikit-learn = "^1.2.0"
-seaborn = "^0.13.0"
 statsmodels = "^0.14.1"
 torch = "^2.3.0"
 torchvision = "^0.18.0"

src/safeds/data/tabular/plotting/_table_plotter.py

@@ -34,12 +34,12 @@ def __init__(self, table: Table):
 
     def box_plots(self) -> Image:
         """
-        Plot a boxplot for every numerical column.
+        Create a box plot for every numerical column.
 
         Returns
         -------
         plot:
-            The plot as an image.
+            The box plot(s) as an image.
 
         Raises
         ------
@@ -52,31 +52,54 @@ def box_plots(self) -> Image:
         >>> table = Table({"a":[1, 2], "b": [3, 42]})
         >>> image = table.plot.box_plots()
         """
-        # TODO: implement using matplotlib and polars
-        import matplotlib.pyplot as plt
-        import seaborn as sns
-
         numerical_table = self._table.remove_non_numeric_columns()
         if numerical_table.column_count == 0:
             raise NonNumericColumnError("This table contains only non-numerical columns.")
-        col_wrap = min(numerical_table.column_count, 3)
+        from math import ceil
 
-        data = numerical_table._lazy_frame.melt(value_vars=numerical_table.column_names).collect()
-        grid = sns.FacetGrid(data, col="variable", col_wrap=col_wrap, sharex=False, sharey=False)
-        with warnings.catch_warnings():
-            warnings.filterwarnings(
-                "ignore",
-                message="Using the boxplot function without specifying `order` is likely to produce an incorrect plot.",
-            )
-            grid.map(sns.boxplot, "variable", "value")
-        grid.set_xlabels("")
-        grid.set_ylabels("")
-        grid.set_titles("{col_name}")
-        for axes in grid.axes.flat:
-            axes.set_xticks([])
-        plt.tight_layout()
-        fig = grid.fig
+        import matplotlib.pyplot as plt
 
+        columns = numerical_table.to_columns()
+        columns = [column._series.drop_nulls() for column in columns]
+        max_width = 3
+        number_of_columns = len(columns) if len(columns) <= max_width else max_width
+        number_of_rows = ceil(len(columns) / number_of_columns)
+
+        fig, axs = plt.subplots(nrows=number_of_rows, ncols=number_of_columns)
+        line = 0
+        for i, column in enumerate(columns):
+            if i % number_of_columns == 0 and i != 0:
+                line += 1
+
+            if number_of_columns == 1:
+                axs.boxplot(
+                    column,
+                    patch_artist=True,
+                    labels=[numerical_table.column_names[i]],
+                )
+                break
+
+            if number_of_rows == 1:
+                axs[i].boxplot(
+                    column,
+                    patch_artist=True,
+                    labels=[numerical_table.column_names[i]],
+                )
+
+            else:
+                axs[line, i % number_of_columns].boxplot(
+                    column,
+                    patch_artist=True,
+                    labels=[numerical_table.column_names[i]],
+                )
+
+        # removes unused ax indices, so there wont be empty plots
+        last_filled_ax_index = len(columns) % number_of_columns
+        for i in range(last_filled_ax_index, number_of_columns):
+            if number_of_rows != 1 and last_filled_ax_index != 0:
+                fig.delaxes(axs[number_of_rows - 1, i])
+
+        fig.tight_layout()
         return _figure_to_image(fig)
 
     def correlation_heatmap(self) -> Image: