Replace fork of Gym-Ignition with upstream release (LocalCache)

[AndrejOrsula]

With the release of Ignition fortress and ever-increasing number of available packages with binaries for ROS 2 rolling, the installation instructions can be easily simplified. I already tested such combination locally together with latest available releases of Gym-Ignition, Stable-Baselines3, PyTorch, ... and everything seems to work fine so far.

Therefore:

• [x] Update README documentation for dependencies and local installation.
• [x] Simplify Dockerfile with binary installation wherever possible

Furthermore, look into this:

• [ ] Try to get rid of custom code inside Gym-Ignition fork and use the upstream release

[AndrejOrsula]

First two tasks were addressed by https://github.com/AndrejOrsula/drl_grasping/pull/94 and https://github.com/AndrejOrsula/drl_grasping/pull/95 (both of which are part of 1.1.0).

---

Regarding the last task for using upstream release of Gym-Ignition, I was able to remove all custom code from the fork with exception of listing models stored in LocalCache (Ignition Fuel) - utils::getLocalCacheModelPaths().

Separating this code to be part of drl_grasping repo, while getting Python bindings, seems like too much effort for seemingly little benefit.

Furthermore, I don't want to open a PR with the current implementation in upstream because LocalCache is now marked as deprecated. I am not exactly sure what is the replacement in Fuel for listing locally cached models that does not involve REST calls to server (which I want to avoid because they take a lot of time).

Therefore, I am keeping this issue opened for now. The optimal solution is to replace LocalCache with something that is not deprecated and then open PR.