changes to topo prediction for new executor

[bfalacerda]

• [ ] only provide Fremen predictions when the edge has been traversed more than x ties
• [x] compare ignoring "ending in wrong node" nav fails when creating predictions vs counting them as successes (current version)

[marc-hanheide]

Here's an idea for @gestom

Can't we use fremen to not only estmate the mean at any given time, but at the same time also try to estimate the variance, and the sample density (each sample is a Dirac, but we could "spread" them out as Gaussians, maybe, to estimate some smaple density) for each time when estimating travel time? This would allow to provide confidence intervals for the estimations (e.g. upper and lower bounds for p=0.05). The topo prediction could, e.g., only trust prediction with a certain confidence, or always be optimistic or pessimistic by choosing the upper or lower bound for its planning? Just an idea, but I could see this as a principled way to incorporate confidence (or almost guarantees) into the system. Not for this year's deployment, but as an idea for future work?

[Jailander]

Second Check is possible using #319 first check would be better done by updating fremenserver to use the recency model I will ask @gestom to start planning for this.