Fix traininstance6 bug

[Opt-Mucca]

Description

I occasionally get fails on traininstance6 and traininstance2 from MIPLIB, and I believe I finally tracked the issue down! In computeDualProof and storeDualInfProof, terms with a coefficient <= epsilon are zeroed out (not relaxed out), which can potentially strengthen the inequality past the point of feasibility. This specifically happens if (1) there's a lot of terms or (2) some of the columns have large bounds. I've then upgraded all the accumulating terms to HighsCDouble and added a new check for zero-ing out terms that val * max(abs(lb), abs(ub)) <= epsilon, i.e., no binaries have a behavioural change. (The isolated bug actually only required the HighsCDouble, but I followed the rabbit-hole for the larger terms first, and I believe the code's better with the change).

I also tried to fix Problem 4 from #2874 while doing this. The issue there is that a cut is potentially scaled by 1e-6 purely for the purpose of trying to extract domain changes. This scales any incorrectness in our cuts by too much, so I've increased the feasibility tolerance check whenever val < 1e-3.

Checklist

• I have read the contributing guidelines
• This PR targets the latest branch
• Tests are passing
• Documentation was updated where relevant
• This PR is not primarily AI-generated (per the AI contributions policy in CONTRIBUTING.md)

Closes #2874

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 73.04%. Comparing base (2b3c167) to head (36075c7).
⚠️ Report is 378 commits behind head on latest.

Additional details and impacted files

@@            Coverage Diff             @@
##           latest    #3223      +/-   ##
==========================================
- Coverage   73.18%   73.04%   -0.15%     
==========================================
  Files         432      445      +13     
  Lines      105505   107770    +2265     
  Branches    16991    17268     +277     
==========================================
+ Hits        77217    78718    +1501     
- Misses      28012    28776     +764     
  Partials      276      276              

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.