[ENHANCEMENT] BrowserGym Env

[burtenshaw]

From working with and deploying the BrowserGym Env for inference and training, I have made these generally useful improvements:

• integrate miniwob server for a 'batteries included' task. IMO, this is useful for beginners.
• improve error handling and environment reset in browsergym
• update the environment example.

[burtenshaw]

cc @sergiopaniego because he's working on TRL examples with this env.