Refine allowance tolerances

[Castavo]

In this PR I offer two changes:

• Change the way we distribute tolerance (today, with too many stops, we can exceed the wanted tolerance, I offer to distribute the tolerance between each section, proportionally to their travel time)
• Change the order in which we compute allowance ranges (today it's from less restrictive to more restrictive, I offer to compute from shortest to largest)

The first commit changes two tests that highlights why I offer to change those behaviors (a test with a short allowance range and a very bug value will fail, and a test with many stops on a path, on top of an allowance accumulates approximation errors)
The next two fix those problems
The last one updates all tests regarding allowances to change the tolerances in assertEquals to be more restrictive

All of those are not atomic, so I'll squash them before merging

EDIT: distributing allowance tolerance according to section relative time makes for a more accurate computation but will make some particular allowances fail (since we'll demand high precision on small ranges (thus a small range with very high allowance value and very high imposed speed will fail). If that's not okay, we may distribute tolerance with a function that favors small values)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (5895b87) 79.55% compared to head (8eb223e) 28.01%.

Additional details and impacted files

@@              Coverage Diff              @@
##                dev    #6404       +/-   ##
=============================================
- Coverage     79.55%   28.01%   -51.54%     
- Complexity     2169     2170        +1     
=============================================
  Files           338     1032      +694     
  Lines         12796   127750   +114954     
  Branches       1775     2603      +828     
=============================================
+ Hits          10180    35793    +25613     
- Misses         1925    90442    +88517     
- Partials        691     1515      +824     

Flag	Coverage Δ
core	78.53% <100.00%> (-0.01%)	⬇️
editoast	75.39% <ø> (?)
front	8.62% <ø> (?)
gateway	2.50% <ø> (?)
railjson_generator	87.25% <ø> (ø)
tests	81.95% <ø> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[axrolld]

New tests look great ! The work with tolerences as well. I'm just not quite sure what to do with the sorting of the ranges that makes the computation sub-optimal in terms of energy consumption. We should probably talk about it.

[axrolld]

📈