Add camera stack for better context switching and safer draw loops

[DragonMoffon]

Currently, only the currently active camera is stored. It would be ideal if more information were stored about every camera activation during a draw loop.

An example of the kind of issue the camera stack fixes: Window draw call

  self.clear()
  self.camera_b.position = self.center
  self.camera_b.use()
  arcade.draw_rect_filled(self.rectangle, (255, 0, 0, 125), self.angle)

  with self.camera_a.activate():
      self.camera_b.position = 100, 100
      arcade.draw_rect_filled(self.rectangle, (255, 0, 0, 255), self.angle)

  arcade.draw_rect_filled(self.rectangle, (255, 0, 0, 125), self.angle)

Current technically correct behaviour

https://github.com/pythonarcade/arcade/assets/86714785/3f4a9644-6479-4b8c-ae98-c6f98f10e880

Improved behaviour post-stack implimentation

https://github.com/pythonarcade/arcade/assets/86714785/1b4bed2f-cc56-4240-99c7-5bcb1e380e45

It also allows for better camera context mangers i.e.

with self.camera:
   ...

rather than

with self.camera.activate():
   ...