The use of `returnfig=True` creates multiple figures when using ipywidgets

Describe the bug
If you provide returnfig=True or use your own axes, updating the ipywidgets produces additional plots.

To Reproduce
this example uses 15 min klines, but that doesn't matter.

@interact(
        start_pos = widgets.FloatSlider(value=0, max=len(df.index) - 1 - 7 * 96, min=0, step=1),
        length    = widgets.FloatSlider(value=7 * 96, max=7 * 96, min=96, step=96),
        
        )
def plot_func(start_pos, length):
    start_pos, length = int(start_pos), int(length)
    ldf = df.iloc[start_pos:start_pos + length]
    print(ldf.index[0], ' - ', ldf.index[-1])
    mpf.plot(ldf, style='yahoo', figsize=(20,5), returnfig=True)
    #print([ str(ax) for ax in axlist])

Expected behavior
When using ipywidgets and updateing the slider, the reloading of the plot_func produces an additional figure each time you update the widget. This shouldn't be. When your don't use returnfig (or don't add ax and volume by your own figure) this behaviour does not accure.

jupyter lab
mplfinance 0.12.10b0

Screenshots

Reproduce the issue in JupyterLab using the shown plot_func with ipywidgets, mplfinance, and returnfig=True. Start by observing figure creation as the sliders update, then trace the plotting entry point and figure handling. Done means repeated widget updates leave only the current plot instead of adding figures.