We have no quantitative metric of irregularity so the best you can hope for are ballpark figures. Also, your conception of regularity (which includes complexity) doesn't exactly match the concept of regularity used in linguistics – Classical Arabic verb conjugation is very complex, but not very irregular in the technical sense. In constructing a word in any language, the typical linguistic analysis is to discern a "base form" for the involved morphemes, which allows you to write rules generating all of the related word forms. That base form is not always some actual word of the language, thus the base form for the verb "be" in Classical Arabic is /kwn/, and kwn is not a possible word of Arabic. There are many regular but complex rules involved in getting a particular word form in Arabic. Given the linguist's viewpoint on regularity, having many rules does not create irregularity: rather, you have irregularity when formatives have to be arbitrarily (lexically) flagged as triggering or not triggering some process.
Again assuming the standard linguistic understanding of regularity / lexicality and the derivation of morphologically complex forms, it would be right to say that gender (be it animacy, sex, or noun-class type) constitutes a kind of irregularity. However, from the popular perspective "learn a word, and heuristically generate related forms based on that", gender is a headache in Germanic but not Spanish, since in learning the citation form of the noun, you can discern the gender most of the time (the real irregularities aren't enough to cause a headache). In some Bantu languages, the difference on this front between linguist's irregularity and popular irregularity is even more striking, since there are over a half-dozen pattern-classes that you just have to learn – but, when you learn the singular (e.g. in Kikuyu) the gender of the noun becomes obvious, since like with Spanish, the basic form of the word that you learn includes the arbitrary gender morpheme that you just have to memorize.
That said, Vietnamese would come in on the most-regular end of the scale, since there are apparently no irregularities or changes in form, whereas there are some idiosyncratic consonant insertions / deletions in Turkish inflection. Dinka comes in on the least-regular end of the scale. Formation of noun plurals is extremely chaotic and arbitrary, and you effectively have to learn singular and plurals. Apparently, outside of a certain core of vocabulary, speakers don't widely agree on singular and plural forms.
