Proper benchmarks for UI #45

Currently we have these tests which are terrible and

• don't actually compare their performance with previous versions of the library
• don't have real examples, just weird toy ones

It might also be nice to design similar UIs in multiple libraries and compare them in terms of performance, but I'm not a huuuge fan of that idea since I want to avoid the whole "omg my ui is so much faster than yours!!" thing (especially because I don't think MLEM.Ui is that much faster than any other ones lmao)