Add tutorial walking through training the visual mesh with webots data

[wongjoel]

This is a practical guide designed to walk a newish member of the team through how to generate a training dataset in Webots, then train the visual mesh on that dataset, then run the newly visual mesh. The main aim is to preserve the knowledge of how to train and run the visual mesh so that it is common knowledge for vision team members. This may also be a good learning tool for people who prefer to learn by following a tutorial rather than reading an explanation of how something works.

This is mostly a placeholder to remind us to finish this after Robocup 2021

Preview: NULL

[ysims]

We need to be careful with the amount of pages we have. I think it is perfectly fine to have a long page with 5-7 headings. Having a sidebar with 50 pages is not so good.

I think this would fit into the page being created in #73, in the section on getting datasets.

If this page is not tutorial-y enough, then we can adjust it or add a section to it.

[LachlanCourt]

@wongjoel What was the decision around this? I think content wise the tutorial is fine although choosing where to put it was the issue?

[wongjoel]

To be honest, I'd kind of forgotten about this: my current thought is that perhaps this project should be geared around revamping the existing guide that uses a dummy dataset. Perhaps the existing guide could be extended to mention how to generate a webots vision dataset (instead of dummy), and also extended to mention how to check if the mesh is actually working.