chore(6993): Add Historical Comparison to UI Startup Metrics

[DDDDDanica]

Description

The Performance Benchmarks section of the MM benchmark PR comment currently has no historical comparison and shows raw numbers with no baseline. The Dapp Page Load benchmark already has a working pattern for this: results are pushed to MetaMask/extension_benchmark_stats on every merge, and the PR comment fetches and displays a comparison against the historical mean.

This PR applies the same historical data pattern to all Performance Benchmarks (Startup, Interaction, and User Journey).

Data collection pipeline:

• Extends benchmark-stats-commit.sh with a ui-startup mode that assembles benchmark results from all platform matrix jobs into a structured JSON payload and commits it to extension_benchmark_stats
• Stores data per-branch (stats/release-12.x/, stats/main/) so PRs compare against the correct release branch baseline
• Startup presets (startupStandardHome, startupPowerUserHome) are stored for all browser/buildType combinations under a pageLoad group; interaction (interactionUserActions) and user journey (userJourney*) presets store chrome-browserify only
• Consolidates all benchmark stats commits (Dapp page load + Performance Benchmarks) into a single store-benchmark-stats job in main.yml, fixing a pre-existing issue where EXTENSION_BENCHMARK_STATS_TOKEN was never wired to the jobs that needed it

Historical fetch module (historical-comparison.ts):

• Fetches ui_startup_data.json for the PR's target branch (GITHUB_BASE_REF)
• When a branch has no data yet (e.g. newly cut release branch), automatically falls back to the most recently populated release branch by semver order
• Produces a HistoricalMeanReference map (benchmark → metric → mean) ready to be consumed by the traffic light / regression detection layer in a follow-up PR

Changelog

CHANGELOG entry: null

Related issues

Fixes: https://github.com/MetaMask/MetaMask-planning/issues/6993

Manual testing steps

1. Go to this page...

Screenshots/Recordings

Before

After

Pre-merge author checklist

• I've followed MetaMask Contributor Docs and MetaMask Extension Coding Standards.
• I've completed the PR template to the best of my ability
• I’ve included tests if applicable
• I’ve documented my code using JSDoc format if applicable
• I’ve applied the right labels on the PR (see labeling guidelines). Not required for external contributors.

Pre-merge reviewer checklist

• I've manually tested the PR (e.g. pull and build branch, run the app, test code being changed).
• I confirm that this PR addresses all acceptance criteria described in the ticket it closes and includes the necessary testing evidence such as recordings and or screenshots.

---

Note

Medium Risk
Modifies CI benchmark workflows and the script that commits results to an external stats repo, so misconfiguration could stop benchmarks from being recorded or break PR benchmark comments, but it does not affect production runtime behavior.

Overview
Adds a per-branch benchmark stats publishing pipeline to MetaMask/extension_benchmark_stats. The updated benchmark-stats-commit.sh now supports dapp-page-load and performance modes, writes to stats/<sanitized-branch>/page_load_data.json and performance_data.json, and aggregates multiple benchmark JSON artifacts (keeping page-load presets across all browser/build combos while storing other presets only for canonical chrome-browserify).

Updates GitHub Actions to rename the dapp page-load workflow/artifact/output file, upload all benchmark JSONs as short-lived artifacts, and introduce a store-benchmark-stats job on main/release/* pushes that downloads artifacts and runs the commit script twice (performance + dapp page-load). Also adds a new TS module + tests for fetching/aggregating historical performance data with release-branch fallback logic, and centralizes benchmark result types in shared/constants/benchmarks (updating imports accordingly).

^{Written by Cursor Bugbot for commit 89c199c. This will update automatically on new commits. Configure here.}

[github-actions]

CLA Signature Action: All authors have signed the CLA. You may need to manually re-run the blocking PR check if it doesn't pass in a few minutes.

[DDDDDanica]

page-load-benchmark.yml no longer touches the stats repo at all. The token is now used directly in the store-benchmark-stats job inside main.yml, which is a regular job (not a reusable workflow) and has direct access to secrets.* without any declaration needed.

[socket-security]

No dependency changes detected. Learn more about Socket for GitHub.

👍 No dependency changes detected in pull request

[metamaskbotv2]

✨ Files requiring CODEOWNER review ✨

🧪 @MetaMask/qa (1 files, +0 -0)

• 📁 test/
  • 📁 e2e/
    • 📁 page-objects/
      • 📁 benchmark/
        • 📄 dapp-page-load-benchmark.ts

---

👨‍🔧 @MetaMask/wallet-integrations (3 files, +3 -3)

• 📁 test/
  • 📁 e2e/
    • 📁 page-objects/
      • 📁 benchmark/
        • 📄 dapp-page-load-benchmark.ts
    • 📁 playwright/
      • 📁 benchmark/
        • 📄 dapp-page-load-benchmark.spec.ts +2 -2
        • 📄 README.md +1 -1

[cursor]

Debug logging dumps entire historical data files

Low Severity

Several [DEBUG]-prefixed console.log statements dump the entire performance_data.json contents via JSON.stringify(currentData, null, 2) and log every individual metric collected. As the stats repo accumulates commits, these dumps will produce extremely large CI log output, making logs harder to read and potentially slowing down the PR comment generation step. The [DEBUG] prefix indicates these were temporary aids, not production logging.

Additional Locations (2)

• development/metamaskbot-build-announce/historical-comparison.ts#L161-L162
• development/metamaskbot-build-announce/historical-comparison.ts#L204-L207

 

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

[cursor]

Subshell exit 0 doesn't stop main script execution

High Severity

The exit 0 inside assemble_performance_data (line 131) only exits the command substitution subshell at line 160 (COMMIT_DATA=$(assemble_performance_data)), not the main script. When no benchmark files are found, COMMIT_DATA is set to an empty string, the script continues past the clone/checkout steps, and then jq --argjson data "" at line 212 fails because an empty string is not valid JSON. This causes an unclean error after git operations instead of the intended graceful skip, and the clone directory is never cleaned up.

Additional Locations (2)

• .github/scripts/benchmark-stats-commit.sh#L159-L160
• .github/scripts/benchmark-stats-commit.sh#L211-L213