docs: update build performance benchmarks (3.0.4)

[github-actions]

Automated build benchmark update for 3.0.4 from workflow run #221.

[greptile-apps]

Greptile Summary

This automated PR updates the build performance benchmarks for version 3.0.4, adding new rows to generated/benchmarks/BUILD-BENCHMARKS.md and syncing the summary table in README.md. The embedded JSON data, per-file metrics, and 50k extrapolations are all internally consistent.

Key changes:

• Native build speed improved by ~50%: 12.3 → 6.2 ms/file
• WASM build speed slightly regressed: 16.3 → 19 ms/file
• Three new build phases introduced: AST Nodes, CFG, and Dataflow — consistent with new analysis capabilities in 3.0.4
• ⚠️ Native no-op rebuild regressed severely: 5ms → 329ms (+6480%) — while the WASM engine holds steady at 7ms. This is a significant real-world regression for users running incremental builds with no changed files, and is worth investigating in the source before this release ships widely.

Confidence Score: 3/5

• Docs are accurate, but benchmarks reveal a severe native no-op rebuild regression that should be resolved before wide release.
• The documentation changes themselves are correct and consistent. However, the benchmarks faithfully record a 6480% regression in native no-op rebuild time (5ms → 329ms) — a real runtime issue reflected in these docs. The PR is safe to merge as documentation, but the underlying performance regression warrants investigation.
• generated/benchmarks/BUILD-BENCHMARKS.md — the incremental rebuild table highlights the native no-op regression worth investigating in source.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Source Files - 177 files] --> B[Parse]
    B --> C[Insert Nodes]
    C --> D[Resolve Imports]
    D --> E[Build Edges]
    E --> F[Structure]
    F --> G[Roles]
    G --> H[AST Nodes]
    H --> I[Complexity]
    I --> J[CFG]
    J --> K[Dataflow]
    K --> L[SQLite DB]

    B -->|Native 413ms| B
    C -->|Native 80ms| C
    D -->|Native 16ms| D
    E -->|Native 143ms| E
    F -->|Native 13ms| F
    G -->|Native 22ms| G
    H -->|Native 300ms| H
    I -->|Native 10ms| I
    J -->|Native 1ms| J
    K -->|Native 76ms| K

    subgraph totals["Build Totals - 3.0.4"]
        N1[Native - 1.1s total - 6.2ms per file]
        W1[WASM - 3.4s total - 19ms per file]
    end

    L --> totals

Loading

_{Last reviewed commit: 129a529}

[greptile-apps]

Significant native no-op rebuild regression

The native no-op rebuild time jumped from 5ms → 329ms (+6480%) in 3.0.4, while the WASM engine remains stable at 7ms. A no-op rebuild should only be checking file hashes against what is stored in the DB — a 65× regression here suggests either a newly introduced DB scan or a lock/connection overhead issue in the native incremental rebuild path.

This is buried in the docs but reflects a real runtime behaviour that will affect users on every incremental build that has no changed files. Worth investigating before shipping 3.0.4 widely.