Calculate stretch threshold based on the optimal route cost

[genadz]

Issue

closes #3239

What's done:
calculate stretch threshold for alternatives based on the optimal route cost. Approximately the following stretch thresholds we will get:

<= 10min -> 2.
20min -> 1.75
30min -> 1.5
1hour -> 1.4
2hours -> 1.3
>= 5hours -> 1.25

In order to approximate these value an quadratic hyperbolic function was chosen (because this is smooth monotonic function).

Tasklist

• Add tests
• Add #fixes with the issue number that this PR addresses
• Update the docs with any new request parameters or changes to behavior described
• Update the changelog
• If you made changes to the lua files, update the taginfo too.

Requirements / Relations

Link any requirements here. Other pull requests this PR is based on?

[genadz]

Here are the tables that demonstrate how number of alternatives changed in this PR. For convenience I split all routes into 4 buckets based on their ETAs.

ETA < 10min

total number of routes = 3392

	master	branch
1-st alternative count	944	1401 (+48%)
2-nd alternative count	345	901 (+160%)

10min < ETA < 30min

total number of routes = 821

	master	branch
1-st alternative count	435	528 (+21%)
2-nd alternative count	254	361 (+42%)

30min < ETA < 1h

total number of routes = 911

	master	branch
1-st alternative count	662	691 (+4%)
2-nd alternative count	490	523 (+7%)

1h < ETA < 5h

total number of routes = 4943

	master	branch
1-st alternative count	3521	3534 (+0.4%)
2-nd alternative count	2312	2320 (+0.3%)

As expected we see significant growth for short routes ( < 30min) and slight changes for medium routes.

[kevinkreiser]

These numbers look great. I assume it doesn't affect p50 that much because it's still really short routes anyway?

[genadz]

Let me share several examples where we started to show alternatives:

ETA < 10min

10min < ETA < 30min

30min < ETA < 1h

1h < ETA < 5h

[genadz]

These numbers look great. I assume it doesn't affect p50 that much because it's still really short routes anyway?

let me check it

[kevinkreiser]

@genadz dont forget to put closes or fixes before the issue number in the description!

[genadz]

Performance

master

5%	10%	50%	90%	95%	99%	100%	
0.0070	0.0101	0.0914	0.3549	0.7473	1.8551	4.2720	
mean: 0.1803

branch

5%	10%	50%	90%	95%	99%	100%	
0.0077	0.0110	0.0920	0.3568	0.7597	1.8818	4.3489	
mean: 0.1820

Timings approximately stayed the same.