Update tidy logic

[dpaddon]

• Updated the min distance/time filtering logic to check each point relative to the last kept point, rather than the previous point in the input list. This previous behaviour could cause too many points to be filtered out.
• For input traces of > maximumPoints coordinates, previously we split the output into multiple features. Now, we randomly sample maximumPoints probes from the trace by removing a random probe until we hit the desired maximumPoints. Note this happens after time/distance filtering.
• Updated the geojson files used for comparison in the tests, so that these conform to the new logic.
• Updated the filtering to allow minimumDistance and minimumTime to be set to 0 (in this case the only effect will be the random sampling if the input length is > maximumPoints). Previously, supplying 0 for these values would cause them to be set to the defaults of 10 and 5 respectively.
• Use an array of probe indices to keep rather than selecting the points directly. This should make it easier in future to allow the function to return properties other than coordTimes, if they are present in the input.

[Valiunia]

because of .filter this is O(n) to remove one point.

I would suggest either use a set for O(1) removals or generate all the points and filter them out in one go

[dpaddon]

If you generate all the points first, you also have to keep track of which ones you have already "removed". By doing it this way, the list is already modified when you move to the next iteration.
This part won't see much heavy usage, as it comes after the time/distance filtering, which naturally reduces the length of the array. I think for now it's not worth optimising too far.

[Valiunia]

why not do something equivalent to

const pointsToRemoveSet = new Set(...) // populate somehow

// at the very end call .filter once
keepIdxs = keepIdxs.filter(element => !pointsToRemoveSet.has(element));

I think this would be much cheaper in the end.

I am just thninking what if there are 1000 points and we need to remove all but 100, we cannot afford to make that many .filter calls

[Valiunia]

also have to keep track of which ones you have already "removed".

not necessarily, I think this can be done efficiently too. For example, shuffle array and take first X points as points to keep/to remove

[tautrimas]

LGTM!