LayoutRestorer for Neu3D is very costly

[TK-21st]

Neu3D does not get disposed of correctly even after refresh, causing memory to build up in browser.

[yiyin]

Also when the JupyterLab browser tab is closed with a lot of neurons loaded in the Neu3D window, the next time JupyterLab is opened, it tries to restore the last state and got stuck. There is a pop up window ask you if you want to keep waiting or clear workspace. Clicking on workspace does not close that pop up window so the session hangs. The JupyterLab tab needs to be closed and reopened for it to work again.

[TK-21st]

should be fixed in cd424a0

In the current implementation, the modelChanged signal triggered by any fbl widget no longer causes the layoutRestorer to automatically store the new state of the widget. Instead, a dirty flag is set for the widget, which is then propogated to the dirty state of the entire Jlab application using the ILabStatus token.

The dirty state is reflected in a few ways, the most noticeable being that in the title bar of the neu3d widget, the close button is a filled circuit when state is dirty. The user can then save the state by either using keyboard shortcut (Cmd+S or Ctrl+S depending on platform) or using the context menu that pops up with right click.

If user tries to close/refresh the window without saving, the entire JLab's application status will be dirty and consequently a pop-up window will be shown: