Use poly as default fitting method everywhere

[CSchoel]

My initial idea to introduce RANSAC as a separate fitting method was to introduce robustness against statistical noise and suboptimal choice for x-axis range. However, I think this convenience just detracts from the work that you need to put into choosing good parameter settings anyway. Worse than that, it can make results non-deterministic and make it hard for first-time users to reproduce results from the literature.

Therefore, I think we can keep RANSAC for experienced users, but switch the defaults to poly everywhere.

DoD

• [ ] Switch to poly fitting everywhere
• [ ] Make a note about this in the changelog