False positive on graphics par() from set_state_inspector()

[joshwlambert]

It's my understanding that simply calling a plotting function changes some of the graphics parameters in the global environment. For example, whether the next plot should be plotted onto a new window (par()$page).

The set_state_inspector() returns warnings when plots are generated when calling devtools::test().

If the plotting function calls on.exit() and restores the graphics parameters this problem is avoided, but on.exit() can lead to issues where the user is trying to modify the graphics parameters globally but they do not persist after the function call (e.g. epiverse-trace/epiparameter#409).

For now I've resolved this issue in {epiparameter} by restoring graphics parameters in each unit test that calls a plotting function, however, this does not scale well and as the title of this issue suggests I feel this is an issue with the state inspector.

Not sure on the best solution, but happy to discuss options.