Browser: Playwright API metrics

[jruales]

No description provided.

[Copilot]

Pull request overview

Adds telemetry for the Integrated Browser run_playwright_code tool to capture Playwright page API usage, success, deferral, duration, and code-size metrics.

Changes:

• Injects telemetry into RunPlaywrightCodeTool and logs completed/failed invocations.
• Adds pageMethodsCalled to Playwright invocation results.
• Wraps Playwright page with a proxy to count method calls and adds a telemetry event helper.

Show a summary per file

File	Description
src/vs/workbench/contrib/browserView/electron-browser/tools/runPlaywrightCodeTool.ts	Logs telemetry for Playwright code tool invocations.
src/vs/platform/browserView/node/playwrightService.ts	Tracks Playwright page API calls via proxy and includes counts in invocation results.
src/vs/platform/browserView/common/playwrightService.ts	Extends invoke result shape with page method call counts.
src/vs/platform/browserView/common/browserViewTelemetry.ts	Defines and emits the new Playwright code telemetry event.

Copilot's findings

Comments suppressed due to low confidence (1)

src/vs/platform/browserView/node/playwrightService.ts:722

• Recursively creating a fresh proxy on every object-property read changes Playwright object identity semantics (for example, page.keyboard === page.keyboard becomes false). Code that caches, compares, or uses nested API objects as map keys can behave differently under this proxy; cache proxies per target/prefix so repeated property reads return the same wrapper.

			if (value !== null && typeof value === 'object') {
				return createPageApiProxy(value as object, methodCalls, `${prefix}${prop}.`, depth + 1);

• Files reviewed: 4/4 changed files
• Comments generated: 4

[Copilot]

Copilot's findings

• Files reviewed: 4/4 changed files
• Comments generated: 1

[kycutler]

Wonder if we should log this from the shared service instead -- so if a call is deferred but the agent never queries for the result, it still gets logged eventually