Running a cell producing a plot for a second time, and the plot fails to display

Nobody has claimed this yet.

## Describe the issue
When using `ipympl` with the `%matplotlib widget` magic, I can successfully display a plot. Rerunning the cell with the plot however, fails to re-display the figure.


the following is visible in the error console


Note: that adding `plt.show()` to the cell and re-running does display a plot **however** the color of the line in the plot is different, suggesting that this is a new (second) matplotlib figure and not the original.

## Versions

```

3.8.10 | packaged by conda-forge | (default, May 10 2021, 22:58:09)
[Clang 11.1.0 ]
ipympl version: 0.9.3
Selected Jupyter core packages...
IPython : 8.10.0
ipykernel : 6.21.2
ipywidgets : 8.0.4
jupyter_client : 8.0.3
jupyter_core : 5.2.0
jupyter_server : 2.3.0
jupyterlab : 3.5.3
nbclient : 0.7.2
nbconvert : 7.2.9
nbformat : 5.7.3
notebook : 6.5.2
qtconsole : not installed
traitlets : 5.9.0
Known nbextensions:
config dir: /Users/stevejpurves/.jupyter/nbconfig
notebook section
jupyter-js-widgets/extension enabled
- Validating: OK
config dir: /Users/stevejpurves/miniconda3/envs/em_recharge/etc/jupyter/nbconfig
notebook section
bqplot/extension enabled
- Validating: OK
ipyevents/extension enabled
- Validating: OK
ipysheet/extension enabled
- Validating: OK
ipytree/extension enabled
- Validating: OK
jupyter-leaflet/extension enabled
- Validating: OK
jupyter-matplotlib/extension enabled
- Validating: OK
jupyter-js-widgets/extension enabled
- Validating: OK
JupyterLab v3.5.3
/Users/stevejpurves/miniconda3/envs/em_recharge/share/jupyter/labextensions
ipytree v0.2.2 enabled OK
jupyterlab_pygments v0.2.2 enabled OK (python, jupyterlab_pygments)
bqplot v0.5.37 enabled OK (python, bqplot)
jupyter-matplotlib v0.11.3 enabled OK
jupyter-leaflet v0.17.2 enabled OK
ipysheet v0.7.0 enabled OK (python, ipysheet)
ipyevents v2.0.1 enabled OK
@jupyter-widgets/jupyterlab-manager v5.0.5 enabled OK (python, jupyterlab_widgets)
```

Start by reproducing the reported `%matplotlib widget` cell rerun in JupyterLab with ipympl, then compare the browser error console and the figure shown on the first and second runs. No source files or tests are named; done means rerunning the cell displays the original figure reliably without requiring `plt.show()` or creating a differently colored second figure.

Written by the indexing model from the issue text.