feat!: Updated KNN index creation interface to support skipping some of source vectors

[glookka]

feat!: Updated KNN index creation interface to support skipping some of source vectors

[github-actions]

Linux debug test results

  8 files    8 suites   11m 24s ⏱️
486 tests 469 ✅ 17 💤 0 ❌
500 runs  483 ✅ 17 💤 0 ❌

Results for commit b56e3d0.

[github-actions]

Linux release test results

  8 files    8 suites   5m 59s ⏱️
486 tests 469 ✅ 17 💤 0 ❌
500 runs  483 ✅ 17 💤 0 ❌

Results for commit b56e3d0.

[github-actions]

Windows test results

  5 files    5 suites   16m 3s ⏱️
469 tests 452 ✅ 17 💤 0 ❌
477 runs  460 ✅ 17 💤 0 ❌

Results for commit b56e3d0.

[Copilot]

Pull Request Overview

This PR updates the KNN index creation API to allow skipping source vectors by passing explicit row IDs.

• Bumps LIB_VERSION to 8
• Extends Builder_i::Train and Builder_i::SetAttr to accept a uRowID parameter
• Modifies HNSWIndexBuilder_c to use caller-provided row IDs and replaces its internal counter with a first-doc flag

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
knn/knn.h	Bumped library version and added uRowID parameter to Builder_i
knn/knn.cpp	Updated HNSW internal builder to accept uRowID, removed counter

Comments suppressed due to low confidence (2)

knn/knn.h:102

• [nitpick] The newly added uRowID parameter on Train and SetAttr should be documented in the header (e.g. a brief doc comment) to explain its role when skipping vectors.

	virtual void	Train ( int iAttr, uint32_t uRowID, const util::Span_T<float> & dData ) = 0;

knn/knn.cpp:394

• [nitpick] There are no tests covering the new behavior where uRowID may be non-sequential or skipped; consider adding unit tests to verify correct indexing when callers supply custom row IDs.

		m_pAlg->addPoint ( (void*)m_dQuantized.data(), (size_t)uRowID );