FIX: Hnsw single thread build (port to 4.0 dev)

[cpegeric]

What type of PR is this?

• API-change
• BUG
• Improvement
• Documentation
• Feature
• Test and CI
• Code Refactoring

Which issue(s) this PR fixes:

issue #24977

What this PR does / why we need it:

fix the bug to keep ThreadsBuild to 1 in all build path. cherry pick from main

[qodo-code-review]

Qodo reviews are paused for this user.

Troubleshooting steps vary by plan Learn more →

On a Teams plan?
Reviews resume once this user has a paid seat and their Git account is linked in Qodo.
Link Git account →

Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center?
These require an Enterprise plan - Contact us
Contact us →

[XuPeng-SH]

I re-checked the current head. I do not see a new logic bug in the single-thread workaround itself, but I still think the regression coverage is below the bar for the exact behavior this PR is fixing.

Right now the only test that actually demonstrates the underlying USearch orphan-node problem is pkg/vectorindex/hnsw/zz_orphan_test.go, and it is still permanently skipped. The normal CI-visible tests do not really pin the fix:

• the build test only requires recall to stay above a broad threshold
• the sync tests mainly assert RunOnce() returns nil
• there is no deterministic assertion that the effective build concurrency is forced to 1 on the relevant paths

That means reverting build.go / sync.go back to multi-threaded add could still slip through CI while silently reintroducing the exact-match reachability bug this PR is supposed to prevent.

Concrete suggestions

1. Add a non-skipped unit test around the actual HNSW build/sync entry points that asserts the effective build/add concurrency is forced to 1 even when config asks for more.
2. Add a deterministic exact-self-match / reachability assertion in a CI-stable test if you have a small enough synthetic dataset for it.
3. At minimum, add focused tests for:
   • GetConcurrencyForSingleThreadBuild()
   • NewHnswBuild(...) choosing single-thread mode
   • NewHnswSync(...) overriding hnsw_threads_build to 1 on the CDC/sync path

I would keep this at request changes until the core workaround is protected by a non-skipped automated regression test, not only by a manual repro file.