workflows: use TinySet/TinyMap in the workflow engine

[mattklein123]

The observation is that we are using a large amount of BTreeSet and BTreeMap which often contains 0 entries, sometimes 1, and very rarely a few. The cost empty construction/destruction of these adds up. This introduces a Set/Map based purely on a Vec and linear scanning.

main::benches::bench_workflow simple:| setup : Setup | setup.simple_workflow()
  Instructions:                        1728|2197                 (-21.3473%) [-1.27141x]
  L1 Hits:                             2642|3443                 (-23.2646%) [-1.30318x]
  LL Hits:                                0|0                    (No change)
  RAM Hits:                             285|345                  (-17.3913%) [-1.21053x]
  Total read+write:                    2927|3788                 (-22.7297%) [-1.29416x]
  Estimated Cycles:                   12617|15518                (-18.6944%) [-1.22993x]
main::benches::bench_workflow many_simple:| setup : Setup | setup.many_simple_workflows()
  Instructions:                       44208|48512                (-8.87203%) [-1.09736x]
  L1 Hits:                            70228|78101                (-10.0805%) [-1.11211x]
  LL Hits:                                7|10                   (-30.0000%) [-1.42857x]
  RAM Hits:                             837|902                  (-7.20621%) [-1.07766x]
  Total read+write:                   71072|79013                (-10.0502%) [-1.11173x]
  Estimated Cycles:                   99558|109721               (-9.26258%) [-1.10208x]

[Copilot]

Pull Request Overview

This PR improves performance by replacing BTreeSet and BTreeMap with custom TinySet and TinyMap implementations optimized for small collections in the workflow engine. The new data structures use Vec-based linear scanning which is more efficient for small sizes (0-5 items) that are common in this context. Benchmarks show significant performance improvements with 8-21% reductions in instructions and cache hits.

• Introduces new TinySet/TinyMap data structures with Vec backing for small collections
• Replaces BTreeSet/BTreeMap usage throughout the workflow engine and related components
• Updates test code to use the new collection types

Reviewed Changes

Copilot reviewed 22 out of 23 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
bd-log-primitives/src/tiny_set.rs	New TinySet and TinyMap implementations with Vec backing
bd-workflows/src/workflow.rs	Replace BTreeMap with TinyMap for logs_to_inject and extraction fields
bd-workflows/src/engine.rs	Update engine to use TinySet/TinyMap for buffer IDs and actions
bd-workflows/src/actions_flush_buffers.rs	Replace BTreeSet with TinySet for buffer action management
bd-workflows/src/workflow_test.rs	Update test expectations to use TinyMap
profiling/src/main.rs	Use TinySet for workflow engine configuration

Comments suppressed due to low confidence (1)

bd-workflows/src/workflow.rs:1

• The method call changed from append to extend which has different semantics. append drains the source collection while extend borrows it. This could cause performance differences or compilation errors if the source is expected to be consumed.

// shared-core - bitdrift's common client/server libraries

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}