Allow more flexible char substitutes

[bt2901]

(see also: pytries/DAWG-Python#4 )

The Russian (е≈ё) and Ukrainian/Belorussian/Rusyn (г≈ґ) orthographies allow some optional letter usage, but it is generally limited to a single "marked" variation of an "unmarked" letter.

However, the orthographies of other Slavic languages require slightly more sophisticated logic for diacritic restoration. Czech has u/ú/ů and e/é/ě; Slovak has o/ó/ô, a/á/ä, and l/ĺ/ľ; Polish has z/ź/ż. The Interslavic constructed language even has three "marked" variations of a single letter: e/ę/ě/ė. (not to mention tonal transcription systems used for Southern Slavic languages)

My fix is relatively straightforward: allow each element of replace_chars to be a list instead of a single char. The performance penalty should be negligible and the backward compatibility should be unaffected.

(EDIT: I hoped everything to be validated by tests via CI, but apparently it doesn't work currently)

[insolor]

Also, it would be nice if you add tests for this functionality

[bt2901]

Added tests. Please note, however, that I hadn't yet tested my changes locally (I have no tools needed to build C++ Python extensions), so someone needs to run tests before merging (ideally, GitHub Actions; I'll look into it later)

[insolor]

CI works, it just needed apporoval to run for a first-time contributor (now I've changed the settings a bit).

Ignore 2.7 tests for now, probably I'll just turn off testing on 2.7.

Other tests really fall after your changes (I didn't check, if that were new tests, or some old tests are broken). You can use github codespaces to run tests not locally.

[insolor]

Ok, I'll merge it today