Add an end-to-end test for ruffle

[Perdu]

This PR adds an end-to-end test, so we can easily test that emulators still work before doing a release. This is a first step that handles ruffle. It could be later extended to other emulators to catch as many errors as possible.

Notes:

• this depends on Dockerfile: move to multi-stage build + fix pcem build #730 as it uses the RUFFLE_VERSION build arg (although it's not really necessary)
• this requires an X server
• I could make this fully headless, but libTAS' headless option gets blocked by this popup: "Some games require old libraries unavailable in recent systems, and libTAS can download them for you, if you allow it (this can be changed at any time in the settings)"
• this downloads a game file that is openly distributed
• libTAS does often segfaults in this as of current version