Enable running multiple torcs

[muupan]

This PR enables running multiple torcs instances, each of which is controlled by different TorcsEnv instances. It would helpful if you run multiple experiments at the same time, or running parallel RL algorithms.

After this PR, if you want to run multiple torcs, just create TorcsEnv instances with different port values.

• Add -title option to torcs so that its window title can be changed
• Use xdotool instead of xte so that it can specify which window to control by its title
• Chage TorcsEnv and Client accordingly
• Remove auto-generated files (This is not related to the subject of the PR. If there's a problem, I'll omit this change.)

[ugo-nama-kun]

Thank you for your great work!
I will immediately pull your request after the check on my ubuntu machine.

[smasoudn]

Please specify relaunch_torcs in line 560 of snakeoil3_gym.py.

Your Client class expect relaunch_torcs in C= Client(p=3101).

[j-lnaddef]

Does it allow to run multiple instances in vision mode?

As far as I understood from #10, they use the image from TORCS' gui, so I guess it is complicated to get the image from multiple instances?

[rtasik]

Hey guys,
can somebody please help me getting multiple clients started? When I try in QuickRace with two scr_servers, the first one is called by python3 snakeoil3_gym.py --port 3101, which connects. In the second console I try python3 snakeoil3_gym.py --port 3102 but then it is not connecting properly and ending up within a loop. Does somebody know how to deal with this?