v4: `errorHandler` receives closed `Page` in browser-based crawlers

[barjin]

Unlike in v3, the v4 BrowserCrawler runs the ContextPipeline cleanup before running the errorHandler. This causes the errorHandler function to receive the closed page every time.

May be intentional, but it's a breaking change nonetheless.

See the recent #3655 - there may be valid use cases for this.