Simplify lifecycle events for frames

[ankur22]

This issue was opened after some discussions in #578.

Basically, the issue was that a test would time out when xk6-browser navigates to google.com and waits for networkIdle. The issue was that we weren't correctly handling the CDP event from the browser.

While working on the fix, it became clear that the lifecycle event code was too complex for what we needed and it might be worth removing.

#593 (comment)

[inancgumus]

The issue was that we weren't correctly handling the CDP event from the browser.

While working on the fix, it became clear that the lifecycle event code was too complex for what we needed and it might be worth removing.

I believe the issue should not be about removing stuff. Instead, finding ways to understand why the existing constructs exist in the first place. For example, in #578, it turned out that the actual problem was about not handling the removal events.

[ankur22]

Research: How do Playwright and Puppeteer do it. Maybe add a Mermaid diagram.

I don't think looking at puppeteer and playwright should be how we determine what's best for xk6-browser. They're a good source of information on why things have been done a certain way in xk6-browser though.

I believe the issue should not be about removing stuff. Instead, #578 (comment) why the existing constructs exist in the first place. For example, in #578, #578 (comment) that the actual problem was about not handling the removal events.

I agree that a suite of tests is what we need now, and then we can refactor the code, whatever the solution maybe as long as the tests pass.

[inancgumus]

I've removed PR #647 from the project board and linked it to this issue. This way, we can reduce the number of issues and can easily track their implementation and tests from a single place. IMO, the project items should mostly be about public-facing issues rather than PRs (not always possible).