[router] remove search parameters from screen.getId()

[marklawlor]

Why

Expo Router's navigate and push functions don't always behave as expected and have overlapping functionality. E.g push wouldn't always push, and navigate's behaviour would change based upon this history in unexpected ways.

For example, lets say we are on the initial URL /page

action	result	visual	notes
navigate('/page?a=1)	/page?a=1	pushed
navigate('/page?b=1')	/page?b=1	pushed
navigate('/page')	/page	rerendered	RN thought this was equal to /page?b=1
goBack	/page?a=1	pop	/page?b=1 skipped

The reason for this behaviour is that RN doesn't preserve the IDs for every screen in the history. It compares them with the current params. The getID function for /page?b=1 and /page are the same, hence this behaviour.

To change how React Navigation does this, we would need to fork the StackNavigator

How

The simple fix is to not include search parameters in the ID generation. This also has the advantage that is now easier to document the differences between push and navigate.

After this PR, Expo Router will have this behaviour

action	result	visual	notes
navigate('/page?a=1)	/page?a=1	rerendered
navigate('/page?b=1')	/page?b=1	rerendered
navigate('/page')	/page	rerendered
goBack	/page	app closes	There is no screen to go back, we only rerender the same screen

action	result	visual	notes
push('/page?a=1)	/page?a=1	pushed
push('/page?b=1')	/page?b=1	pushed
push('/page')	/page	pushed
goBack()	/page?b=1	popped

Test Plan

I've updated the unit tests and also added a new set to test stack unwinding.

Checklist

• Documentation is up to date to reflect these changes (eg: https://docs.expo.dev and README.md).
• Conforms with the Documentation Writing Style Guide
• This diff will work correctly for npx expo prebuild & EAS Build (eg: updated a module plugin).

[EvanBacon]

Should check the expected history too

[marklawlor]

Blocked by #27307

I want to ensure we have the push behaviour correct before fixing the navigate behaviour. These PRs also modify the same tests, but #27307 adds more detail