Optimize channel computation

[Technici4n]

• Optimize the channel computation to make it go from O(N^2) to O(N).
• Cleanup related code a little bit. Most importantly, choose the same meaning for usedChannels vs lastUsedChannels for both GridConnection and GridNode.
• Remove booting delay. Note that the API still supports it, and that we might want to add back some sort of delay in the future if necessary.
• Remove pathfindingStepsPerTick config option.

[shartte]

I think you mean it goes from O(N^2) to O(N) I hoper

[bk5115545]

Since the method of jumping the path calculation is essentially what timsort does, I think this will be closer to O(N * N/(average jump size for the current network) + (average path length to controller * number of channel consumers)).

Time contributed from pathing: N * N/(average jump size for the current network). This is an optimization of BFS which is normally N * (edge count).
Time contributed from channel assignment: (average path length to controller) * (number of channel consumers)

Still a crazy improvement over the current implementation but I think if you invented an O(N) graph pathing algorithm then you'd solve the travelling salesman problem so I'm hesitant to believe this is O(N)

[Technici4n]

I think you mean it goes from O(N^2) to O(N) I hoper

Yes 😄

[Technici4n]

The new algorithm is linear time because there are at most 2 jumps. (In theory. I have to double-check that this is true in practice.)

As a side note, BFS and DFS are both linear time traversal algorithms. And BFS can find shortest paths in linear time, the catch is that it only works with unweighted edges.

[bk5115545]

Is this something that can be backported to 1.20 (I don't know your guidelines on versioning)?

I'd really like to get it into Monifactory and I can do the backport work/make a PR if you're alright with that

[shartte]

@bk5115545 Backported as 15.3.0-beta

[bk5115545]

You're amazing. Thank you!