[Feature] ⚡ Parallel Scenario Execution with Browser Pool

[khaledbourghida]

Problem Statement

Scenarios currently execute sequentially in a single browser context. A 10-scenario suite takes 10x the time of a single scenario. This is fine for demos but unusable for real test suites of 30+ scenarios.

Proposed Solution

Replace sequential execution with a Browser Pool that runs N scenarios concurrently using isolated browser contexts.

Architecture

Browser Pool (max N concurrent, default 4) ├─ Context A → Scenario 1 ├─ Context B → Scenario 2 ├─ Context C → Scenario 3 └─ Context D → Scenario 4 ↓ When one finishes, pull next scenario from queue

Implementation Notes

• browser.js already uses a sessions Map with getBrowserForSession(). Extend this into a BrowserPool class.
• Use browser.newContext() per scenario (cheap) — NOT a new browser per scenario (expensive).
• Use Promise.allSettled with concurrency control (e.g. p-limit or manual semaphore).
• Critical: Each context needs its own selectorDict copy. The learned_selectors SQLite cache is already keyed by page_url + original_selector and uses WAL mode, so concurrent writes are safe.
• Add pool.maxConcurrency to settings.json (timeouts section or new execution section).

Which Agent / Component Does This Affect?

• Browser Executor
• Code Analyst
• Route Navigator
• Selector Expert
• Scenario Writer
• Script Writer
• Verifier / Step Fixer
• Cache / SQLite layer
• UI / Client
• Configuration / Settings
• Other: server/browser.js — BrowserPool class