refactor: move zero width replacement to a function

[Nerixyz]

Refactors and tests the Unicode zero-width-joiner replacement. The replacement is now in a helper function (so we can test it) and the individual components use Unicode string literals. (From #5588).

[pajlada]

I would prefer if this was split into two PRs (or at the very least commits with tests being allowed to fully run)

1. Moving the logic to a function, and adding tests for it
2. Change way we encode the characters