Fix SDK->simruntime mapping assumptions in test runners

[aaronsky]

Redux of #2675.

Refactors simulator_creator.py to discover a recent simruntime when none is explicitly declared, and removes the SDK version fallback from the test runner rule implementations. This assumption is problematic in recent times whenever Apple pushes a new simruntime OTA for an existing Xcode, or when the simruntime identifier does not align with the simruntime version and consequently the SDK version (such as OS patch releases).

[adincebic]

Looks good at first glance. Are identifier and version enough to match the runtime or do we have to take platform into account as well? Perhaps identifier is enough?

[aaronsky]

The runtime identifier is in the format com.apple.CoreSimulator.SimRuntime.iOS-26-3 and should be enough. SDK version and OS version should be okay given an assumption (or eventual parameterization) of a platform. So much of the test runner logic assumes iOS, which I would like to fix in a subsequent PR.

The only thing I'm not sure about is whether falling back on runtime-match is sufficient – for example, I bet if you downloaded Xcode 26.3 and the iOS 26.3 simruntime now, your runtime-match would be incorrect. However, in the case of CI's setup, that's clearly enough. 🤷 Maybe I should still incorporate a basic "does the SDK version match the runtime version" check into the loop?

This also doesn't resolve the issues in this area with actool et al, which still assume the SDK and simruntime (which actool leverages) are the same.

[luispadron]

Thank you!