Refactor #system_connect without yield

[straight-shoota]

Socket#system_connect yields in case of an error. I understand the intention to avoid actually raising an exception for efficient error handling in case multiple targets are being tried (for example with Addrinfo.resolve).
But I’m not sure what’s the motivation for yielding instead of returning the Exception. The yield always happens just right before the method returns. This is a simple refactor of an internal method to make it simpler and easier to reason about.

The public method Socket#connect keeps the yielding behaviour to not break the API.

This cleans up the internal interface a in preparation of a bigger refactoring of the event loop (#10766).

[ysbaddaden]

Yes, the initial idea (#3750) was to avoid raising/catching an exception when trying multiple IP to connect to. I suppose the yield crept up to the internal calls.

[straight-shoota]

I suppose yielding could make some sense if the exception was allocated on the stack to avoid heap allocations. Then a caller who wants to raise, would need to clone it to the heap first. That seems a bit brittle, though.

[straight-shoota]

FTR: I was wondering whether the yield setup could have any effect on the strack trace. But it seems it does not. The strack trace is exactly the same with or without this change (except for shifted line numbers of course).