i/6188: The uid() helper should be a lot faster

[oleq]

Suggested merge commit message (convention)

Internal: The uid() helper should be a lot faster. Closes ckeditor/ckeditor5#6188.

[coveralls]

Coverage remained the same at 100.0% when pulling 2fa4510 on i/6188 into e63b7b7 on master.

[Reinmar]

Since we were working on it in the context of ckeditor/ckeditor5#6161 (comment) I'd like to know its effect on that scenario. Could you compare before/after performance of loading that UI?

[oleq]

Nothing spectacular. Around 35ms on total 350ms when switching categories from Arrows->All (~10%).

Then I checked the http://localhost:8125/ckeditor5/tests/manual/performance/richeditor.html test and I created the editor on button click. The difference is 965ms vs 895ms (around 7%).

It's not like the editor is blazingly fast now but this is something.