Launching the GPU instance anytime when checking the performance guarantee off

[icoxfog417]

Is your feature request related to a problem? Please describe.
As a student, GPU allocation has more priority than performance when running the Jupyter Notebook materials that include models such as GAN/Transformer in their lecture. The medium performance GPU is enough but running on CPU causes the delay of progress of the time limited lecture.

Describe the solution you'd like
The students can run GPU instance anytime if they turn on the performance not guarantee GPU checkbox when launching the GPU instance.

Describe alternatives you've considered
The students give up using Studio Lab and select other free notebook service that they can use GPU.

Additional context

This issue is from #92