Translit ru exceptions

[markalex2209]

A bunch of minor changes to ru transliteration checker:

• Lowered weigh of difference between pairs х/г and а/я
• Made Эй an expected transliteration for Ei in the beginning of word
• Moved check for šja/ā to actually work
• Added soft sign between soft consonant followed by another consonant (other then the same one, like ņņ)
• Added soft sign between consonant and group j + vowel: mju -> мью not мю

Additionally, lowered "good enough" threshold to 0.5: I'm working my way through transliterations, and I hope to address most of them that actually differ in at least a letter.

[markalex2209]

Let's wait a bit. I'm also checking combinations of consonant + j + vowel, like mju. I think they all need to be transliterated with soft sign between consonant and vowel, but I need to check.

[markalex2209]

Checked, confirmed. Happy with results. Ready for review

[HellMapGoesCoding]

Just as a reflection. I originally wanted to add an existing package for transliterations and I tried a couple. But I immediately realized that none of them handle Latvian language and all those š and ž letters with diacritics don't get transliterated at all. So I started making my own. But when I got to the point having to think about consonants/vowels and combinations, I just stopped caring... So that's why I also added the approximate comparison.

[markalex2209]

I get you. And I think not using library here is wise, since without it we are able to tweak process as we like (For example, combination of jo might require attention in future). And I got that since you are not native speaker, this might be a bit too much. I'm glad to help were I can.

Regarding word distance - this is a correct call, because manually comparing strings will be a lot of useless headache. And I personally don't care if the most appropriate transliteration would be Саиета but now it is Сайета (or vice versa), etc.

So, I'd say your choices played out very well here.