Fix clique size and numerical issue bugs

[akifcorduk]

Small cliques are reduced to pairwise conflicts in adj list format and the limit was 512. For some problems, this is too agressive and creates large amount of adj list entries that exceed int32 size. This PR fixes it by reducing the adj list storage threshold to 64.

This PR also fixes a roundoff error that is revealed after extraction of cliques from mixed rows were instroduced. We check with tolerance now.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 3e14be5e-ff02-42e2-89ee-a4939ed055a8

📥 Commits

Reviewing files that changed from the base of the PR and between 013f87c and 13fd597.

📒 Files selected for processing (2)

• cpp/src/mip_heuristics/presolve/conflict_graph/clique_table.cu
• cpp/tests/mip/cuts_test.cu

🚧 Files skipped from review as they are similar to previous changes (1)

• cpp/src/mip_heuristics/presolve/conflict_graph/clique_table.cu

---

📝 Walkthrough

Walkthrough

Changes

Clique graph updates

Layer / File(s)	Summary
CSR storage and clique configuration cpp/src/mip_heuristics/presolve/conflict_graph/clique_table.cuh	CSR offsets and compaction counters use int64_t, while the default minimum clique demotion size changes from 512 to 64.
Tolerance-aware clique construction cpp/src/mip_heuristics/presolve/conflict_graph/clique_table.cu, cpp/tests/mip/cuts_test.cu	Clique construction applies absolute RHS tolerance during initial checks and includes a mixed-row roundoff regression test preventing a spurious conflict edge.
Additional-clique tolerance boundary cpp/src/mip_heuristics/presolve/conflict_graph/clique_table.cu, cpp/tests/mip/cuts_test.cu	Additional-clique search uses upper_bound and tolerance-adjusted validation, with a regression test for tolerance-boundary conflicts.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Suggested labels: mip

Suggested reviewers: kh4ster, nguidotti, mlubin

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately summarizes the main fixes: clique size handling and numerical tolerance issues.
Description check	✅ Passed	The description is clearly related to the threshold reduction and tolerance-based roundoff fix in the changeset.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@cpp/src/mip_heuristics/presolve/conflict_graph/clique_table.cu`:
- Around line 33-34: The additional-clique path must use the same strict
tolerance boundary as the initial check: update the logic around the lower_bound
handling to reject candidate sums equal to rhs_with_tolerance and proceed only
when the sum is strictly greater. Add a regression test that includes a valid
base clique and reaches the additional-clique path, verifying equality at the
tolerance boundary does not reintroduce the conflict.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 470cbd38-3ffb-4813-97a8-be8db5576eb4

📥 Commits

Reviewing files that changed from the base of the PR and between e4e1aad and 013f87c.

📒 Files selected for processing (3)

• cpp/src/mip_heuristics/presolve/conflict_graph/clique_table.cu
• cpp/src/mip_heuristics/presolve/conflict_graph/clique_table.cuh
• cpp/tests/mip/cuts_test.cu

[aliceb-nv]

Good forward thinking for the int64_t for offsets :) We should probably go over the whole codebase someday to ensure all CSRs use 64bit for the offsets. Perhaps a off_t typedef or such and agentic checks

[github-actions]

CI Test Summary

✅ All 31 test job(s) passed.

[akifcorduk]

/merge