feat: per-stage indexing timings from the optimizations endpoint

[tellet-q]

Indexing was a single number — wait_secs — so a slow index build told you nothing about what was slow. Qdrant already tracks the breakdown; it just isn't exposed over gRPC. This reads it and reports it. No new subcommand, no new workflow. bfb upload and the legacy run already wait for a green index before exiting; that wait now also reports what it cost:

  bfb upload --file config.yaml -n 1M --json results.json
                                                                                                                                                                                                                                              
  Index ready in 20.255 seconds
  --- Optimization stages (1 completed, 18.068 s total) ---
  vector_index/main_graph:      16.663360 s total over 1 run(s), max 16.663360 s
  copy_data:                     0.336660 s total over 1 run(s), max 0.336660 s
  payload_index/keyword:color:   0.133206 s total over 1 run(s), max 0.133206 s

Attribution is baseline-diffed. The server exposes a rolling window of the last 16 completed optimizations, not the ones belonging to your run. So bfb snapshots the completed UUIDs before the phase and excludes them afterwards — otherwise a run would cheerfully report an earlier one's work as its own.

All Submissions:

• Contributions should target the dev branch. Did you create your branch from dev?
• Have you followed the guidelines in our Contributing document?
• Have you checked to ensure there aren't other open Pull Requests for the same update/change?

[Copilot]

Pull request overview

Adds per-stage indexing/optimization timing attribution to the existing “wait for green index” phase by reading Qdrant’s REST-only /collections/{collection}/optimizations endpoint, baseline-diffing completed optimization UUIDs, printing a stage breakdown, and persisting the report into the JSON results document.

Changes:

• Introduces a REST-backed optimizations module that fetches and aggregates optimization progress-tree stage timings with baseline exclusion.
• Extends results.index to optionally include optimizations (per-stage aggregates) alongside wait_secs.
• Adds --skip-server-stats to disable the REST-only optimization breakdown and updates docs/fixtures accordingly.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
tests/fixtures/optimizations_completed.json	Adds a real-server JSON fixture used to validate parsing and stage aggregation.
src/results.rs	Extends IndexPhase to optionally store the optimization-stage report in JSON output.
src/optimizations.rs	New module to fetch, baseline-diff, aggregate, and print per-stage optimization timings from Qdrant REST.
src/main.rs	Captures optimization baseline before upload, waits for green index, then fetches/prints/stores stage breakdown.
src/args/mod.rs	Adds --skip-server-stats flag to disable REST-only extras.
README.md	Documents the new per-stage indexing timing output and JSON field location.
DEVELOPMENT.md	Updates file layout documentation to include the new optimizations.rs module.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.