Use custom struct instead of std::pair in QueryHeap

[SiarheiFedartsou]

• improves readability
• seems to be a microoptimisation

Benchmark Results

Benchmark	Base	PR
alias	aliased u32: 1087.89 plain u32: 1085.74 aliased double: 957.784 plain double: 954.912	aliased u32: 1094.16 plain u32: 1143.48 aliased double: 1181.03 plain double: 1173.1
json-render	String: 6.5949ms Stringstream: 9.33359ms Vector: 6.8915ms	String: 6.56282ms Stringstream: 9.17836ms Vector: 6.80801ms
match_ch	Default radius: 4.46775ms/req at 82 coordinate 0.0544847ms/coordinate Radius 5m: 4.4405ms/req at 82 coordinate 0.0541524ms/coordinate Radius 10m: 15.1805ms/req at 82 coordinate 0.185128ms/coordinate Radius 15m: 37.1138ms/req at 82 coordinate 0.452607ms/coordinate Radius 30m: 317.423ms/req at 82 coordinate 3.87101ms/coordinate	Default radius: 4.42469ms/req at 82 coordinate 0.0539597ms/coordinate Radius 5m: 4.38681ms/req at 82 coordinate 0.0534976ms/coordinate Radius 10m: 15.0091ms/req at 82 coordinate 0.183038ms/coordinate Radius 15m: 36.6437ms/req at 82 coordinate 0.446875ms/coordinate Radius 30m: 312.646ms/req at 82 coordinate 3.81276ms/coordinate
match_mld	Default radius: 3.34427ms/req at 82 coordinate 0.0407838ms/coordinate Radius 5m: 2.79952ms/req at 82 coordinate 0.0341405ms/coordinate Radius 10m: 10.3098ms/req at 82 coordinate 0.125729ms/coordinate Radius 15m: 26.4945ms/req at 82 coordinate 0.323104ms/coordinate Radius 30m: 310.665ms/req at 82 coordinate 3.7886ms/coordinate	Default radius: 2.78484ms/req at 82 coordinate 0.0339614ms/coordinate Radius 5m: 2.7996ms/req at 82 coordinate 0.0341415ms/coordinate Radius 10m: 10.1723ms/req at 82 coordinate 0.124052ms/coordinate Radius 15m: 26.0148ms/req at 82 coordinate 0.317254ms/coordinate Radius 30m: 304.017ms/req at 82 coordinate 3.70753ms/coordinate
packedvector	random write: std::vector 9858.82 ms util::packed_vector 82186.3 ms slowdown: 8.33632 random read: std::vector 8513.61 ms util::packed_vector 33472.8 ms slowdown: 3.93168	random write: std::vector 11236.2 ms util::packed_vector 73886.6 ms slowdown: 6.57575 random read: std::vector 11103.5 ms util::packed_vector 30234 ms slowdown: 2.72293
route_ch	1000 routes, 3 coordinates, no alternatives, overview=full, steps=true 509.659ms 0.509659ms/req 1000 routes, 2 coordinates, no alternatives, overview=full, steps=true 351.646ms 0.351646ms/req 1000 routes, 2 coordinates, 3 alternatives, overview=full, steps=true 628.2ms 0.6282ms/req 1000 routes, 3 coordinates, no alternatives, overview=false, steps=false 152.53ms 0.15253ms/req 1000 routes, 2 coordinates, no alternatives, overview=false, steps=false 97.4741ms 0.0974741ms/req 1000 routes, 2 coordinates, 3 alternatives, overview=false, steps=false 132.57ms 0.13257ms/req 1000 routes, 3 coordinates, no alternatives, overview=false, steps=false, radius=750 151.428ms 0.151428ms/req 1000 routes, 2 coordinates, no alternatives, overview=false, steps=false, radius=750 97.0949ms 0.0970949ms/req 1000 routes, 2 coordinates, 3 alternatives, overview=false, steps=false, radius=750 132.039ms 0.132039ms/req	1000 routes, 3 coordinates, no alternatives, overview=full, steps=true 511.645ms 0.511645ms/req 1000 routes, 2 coordinates, no alternatives, overview=full, steps=true 353.033ms 0.353033ms/req 1000 routes, 2 coordinates, 3 alternatives, overview=full, steps=true 630.633ms 0.630633ms/req 1000 routes, 3 coordinates, no alternatives, overview=false, steps=false 149.976ms 0.149976ms/req 1000 routes, 2 coordinates, no alternatives, overview=false, steps=false 96.5049ms 0.0965049ms/req 1000 routes, 2 coordinates, 3 alternatives, overview=false, steps=false 132.923ms 0.132923ms/req 1000 routes, 3 coordinates, no alternatives, overview=false, steps=false, radius=750 149.583ms 0.149583ms/req 1000 routes, 2 coordinates, no alternatives, overview=false, steps=false, radius=750 96.4442ms 0.0964442ms/req 1000 routes, 2 coordinates, 3 alternatives, overview=false, steps=false, radius=750 131.794ms 0.131794ms/req
route_mld	1000 routes, 3 coordinates, no alternatives, overview=full, steps=true 643.621ms 0.643621ms/req 1000 routes, 2 coordinates, no alternatives, overview=full, steps=true 441.984ms 0.441984ms/req 1000 routes, 2 coordinates, 3 alternatives, overview=full, steps=true 818.912ms 0.818912ms/req 1000 routes, 3 coordinates, no alternatives, overview=false, steps=false 280.732ms 0.280732ms/req 1000 routes, 2 coordinates, no alternatives, overview=false, steps=false 164.269ms 0.164269ms/req 1000 routes, 2 coordinates, 3 alternatives, overview=false, steps=false 300.394ms 0.300394ms/req 1000 routes, 3 coordinates, no alternatives, overview=false, steps=false, radius=750 272.22ms 0.27222ms/req 1000 routes, 2 coordinates, no alternatives, overview=false, steps=false, radius=750 161.836ms 0.161836ms/req 1000 routes, 2 coordinates, 3 alternatives, overview=false, steps=false, radius=750 293.598ms 0.293598ms/req	1000 routes, 3 coordinates, no alternatives, overview=full, steps=true 640.204ms 0.640204ms/req 1000 routes, 2 coordinates, no alternatives, overview=full, steps=true 436.477ms 0.436477ms/req 1000 routes, 2 coordinates, 3 alternatives, overview=full, steps=true 815.515ms 0.815515ms/req 1000 routes, 3 coordinates, no alternatives, overview=false, steps=false 268.377ms 0.268377ms/req 1000 routes, 2 coordinates, no alternatives, overview=false, steps=false 159.875ms 0.159875ms/req 1000 routes, 2 coordinates, 3 alternatives, overview=false, steps=false 287.392ms 0.287392ms/req 1000 routes, 3 coordinates, no alternatives, overview=false, steps=false, radius=750 261.197ms 0.261197ms/req 1000 routes, 2 coordinates, no alternatives, overview=false, steps=false, radius=750 161.217ms 0.161217ms/req 1000 routes, 2 coordinates, 3 alternatives, overview=false, steps=false, radius=750 287.302ms 0.287302ms/req
rtree	1 result: 207.75ms -> 0.020775 ms/query 10 results: 243.355ms -> 0.0243355 ms/query	1 result: 207.172ms -> 0.0207172 ms/query 10 results: 243.418ms -> 0.0243418 ms/query

[DennisOSRM]

Nice readability improvement

[SiarheiFedartsou]

Nice readability improvement

I hope it is not only readability :) locally I have slight performance improvement on some of benchmarks. It can be confusing, but there are indeed some reasons why std::pair can be slower than raw struct:
https://danlark.org/2020/04/13/why-is-stdpair-broken/
https://www.reddit.com/r/cpp/comments/ar4ghs/stdpair_disappointing_performance/

But tbh I am not 100% sure yet it is the case here.

[SiarheiFedartsou]

e indeed some reasons why std::pair can be slower than raw struct: https://danlark.org/2020/04/13/why-is-stdpair-broken/ https://www.reddit.com/r/cpp/comments/ar4ghs/stdpair_disappointing_performance/

But tbh I am not 100% sure yet it is the case here.

Well, according to our CI benchmark it seems to be indeed a microoptimisation, but hard to say for sure as improvement is really small. Also weird that I see it only for MLD(UPDATE: after re-run there is slight improvement on CH as well, but it is hard to say if it is statistically significant). But as this is nice as readability only thing I propose to merge it anyway.

[SiarheiFedartsou]

It seems it guarantees that we will get keys with the same weight in insertion order and I added it only to preserve guarantees we had before with std::pair, but is it really needed? 🤔