Add InsertResult::SuccessUpdate

[Jack Moffitt (metajack)]

Garnet protocol requires different results for VADD depending on whether it was an insert or an update. This adds an InsertResult::SuccessUpdate variant to flag that case. Also, it turned out updates weren't actually implemented at all, so this also fixes that issue.

Because quantization backfill may interact with updates, a new range mutex is introduced to allow quantization backfill and updates to lock ranges for changes so they don't stomp on each other.

[Copilot]

Copilot review overview

🟡 Changes recommended

Update detection can corrupt ID mappings after failed deletion and remains unsafe for concurrent inserts of the same external ID.

Review effort: Balanced
Findings: 1 · 1

Open (2)

• Concurrent inserts can duplicate external ID mappings · New
• Update failures are logged with the wrong context term · New

What changed in this PR

Adds vector-update support to the Garnet provider and exposes updates through the FFI result contract.

Changes:

• Reuses internal IDs for updates and adds SuccessUpdate.
• Coordinates quantization backfill with vector updates.
• Adds update/backfill tests, documentation, and a major version bump.

File	Description
Cargo.lock	Updates the crate version.
diskann-garnet/​Cargo.toml	Bumps version to 6.0.0.
diskann-garnet/​diskann-garnet.nuspec	Aligns NuGet package version.
diskann-garnet/​docs/​ffi-design.rs	Documents update result semantics.
diskann-garnet/​src/​ffi_recall_tests.rs	Simplifies unaligned ID decoding.
diskann-garnet/​src/​ffi_tests.rs	Tests update results and ID reuse.
diskann-garnet/​src/​fsm.rs	Adds guards for existing IDs.
diskann-garnet/​src/​garnet.rs	Tracks update state in contexts.
diskann-garnet/​src/​lib.rs	Exposes SuccessUpdate through FFI.
diskann-garnet/​src/​provider.rs	Implements updates and range synchronization.
diskann-garnet/​src/​test_utils.rs	Tracks internal-map reads in tests.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Codecov Comments Bot (codecov-commenter)]

Codecov Report

❌ Patch coverage is 97.01087% with 11 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.60%. Comparing base (fcf9053) to head (e5cf957).

Files with missing lines	Patch %	Lines
diskann-garnet/src/garnet.rs	52.17%	11 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1449      +/-   ##
==========================================
+ Coverage   90.58%   90.60%   +0.01%     
==========================================
  Files         568      568              
  Lines      112680   113023     +343     
==========================================
+ Hits       102069   102401     +332     
- Misses      10611    10622      +11     

Flag	Coverage Δ
miri	90.60% <97.01%> (+0.01%)	⬆️
unittests	90.36% <97.01%> (+0.01%)	⬆️

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

Files with missing lines	Coverage Δ
diskann-garnet/src/fsm.rs	94.52% <100.00%> (+0.07%)	⬆️
diskann-garnet/src/lib.rs	94.39% <100.00%> (ø)
diskann-garnet/src/provider.rs	82.45% <100.00%> (+3.40%)	⬆️
diskann-garnet/src/test_utils.rs	98.61% <100.00%> (+0.03%)	⬆️
diskann-garnet/src/garnet.rs	93.31% <52.17%> (-2.17%)	⬇️

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

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

[Harsha Simhadri (harsha-simhadri)]

IT would be nice to validate this PR on runbooks with replace operations. Thanks.

[Jack Moffitt (metajack)]

Results from wikipedia-1M using wikipedia-1M_expiration_time_replace_only_runbook.yaml. l_build=300 and maxR=48.

### Recall@100

 Positive differences mean Garnet is higher.

 ┌──────────┬─────────────────────┬───────────────┬──────────────────┬────────────────────────────────────┐
 │ L-search │Baseline             │ Garnet, final │ Final difference │ Mean difference across checkpoints │
 ├──────────┼─────────────────────┼───────────────┼──────────────────┼────────────────────────────────────┤
 │ 100      │ 88.1172%            │ 88.3144%      │ +0.1972 pp       │ +0.2180 pp                         │
 ├──────────┼─────────────────────┼───────────────┼──────────────────┼────────────────────────────────────┤
 │ 192      │ 94.0728%            │ 94.0898%      │ +0.0170 pp       │ +0.0222 pp                         │
 ├──────────┼─────────────────────┼───────────────┼──────────────────┼────────────────────────────────────┤
 │ 300      │ 96.6184%            │ 96.5974%      │ −0.0210 pp       │ −0.0092 pp                         │
 └──────────┴─────────────────────┴───────────────┴──────────────────┴────────────────────────────────────┘