Refactor population structure modules with PCA and NJT methods

[alimanfoo]

The PCA module is hard, and PCAs are just super hard to interpret, lots of people get confused. Genetic distance and neighbour-joining trees are a more intuitive and easier to interpret method.

Propose to:

• Create a training module on population structure with neighbour-joining trees, to replace the current PCA module in the entry-level training course.
• Migrate the PCA training to an advanced-level module.