Limit on cost

[t2gran]

Provide a limit on cost. There is several ways to do this:

1. We can set a maxLimit by calculating maxLimit = a * minCost + b. The a and b are constants passed in as parameters. The minCost is the is the best cost found at the destination - dynamiclly updated when new destinations arrivals are found.
2. We can specify a delta constant that is used in every stop for comparing cost. For every pareto compare operation the new cost must be better than an existing value by a delta to be added to the set.So let say we use walkingDistance as cost. An arrival at a given stop arrive with 500 meters walking. Then later a new arrival arrive with 490 meters walking. The later arrival is pareto optimal if delta is 0 (<10) and not optimal (rejected) if delta is > 10.

The first alternative prune of the rim of the search, while the second start pruning on every stop.

In principle these optimizations can be applied to other criteria as well.

[t2gran]

A simple test of this uses a reverse, no-wait search to calculate min (transfers, time, and cost) and then uses these three to create an optimistic estimate for the destination arrival in the multi-criteria search.

For my 27 test cases searching from 0800 and 12 hour window the search time go down from 3000 ms to 900 ms - the reverse search uses on average 160 ms. With a 4 hour time window the search window only improve from 560ms to 500ms. The reverse search took 160 ms.

[t2gran]

Implemented using a heuristic calculation and limit by by testing if the arrival qualify at the destination. See commit 1ecf8add and 5250f26b.