feat: LLM Cost Tracking and Budgeting feature in passmark

[devhimanshuu]

🌟 Summary

Introduces an enterprise-grade LLM Cost Tracking & Budgeting system. This prevents "bill shock" by strictly monitoring API costs across all parallel test workers and halting execution if a user-defined budget is exceeded.

❓ Why is this needed?

AI-driven tests can accumulate high API costs rapidly, especially during loops or parallel test execution in CI/CD. Teams need strict, observable cost controls to confidently run Passmark in production pipelines without fear of runaway expenses.

🛠️ Key Features

• Distributed Budget Enforcer: Added maxCostPerRun to configure(). If the budget is breached, it safely aborts the suite with a BudgetExceededError.
• Global Worker Sync (Redis): Uses Redis (INCRBYFLOAT) with a 24-hour TTL to precisely aggregate costs across all parallel Playwright workers, ensuring true global limits.
• Custom Pricing Overrides: Added a pricing config option to define custom token rates (useful for enterprise gateways or self-hosted models).
• OpenTelemetry Observability: Automatically attaches token counts and cost metrics (ai.usage.cost) to active OpenTelemetry traces for rich Datadog/Axiom dashboarding.
• CI/CD Artifacts: Generates a passmark-cost-<pid>.json summary in the uploads directory and logs the total cost upon process exit.
• Universal Tracking: Intercepts Vercel AI SDK usage and raw OpenAI Responses API (CUA mode) calls universally.

💻 Usage Example

import { configure } from "passmark";

configure({
  maxCostPerRun: 2.00, // Halts the test suite if AI cost exceeds $2.00 globally
  pricing: {
    // Optional: override default pricing
    "my-internal-model": { input: 0.10, output: 0.20 }
  }
});