When Files with a cyrillic characters were used, the relativepath and path component
would hold differently compositions(codepoints) of that character, which led to a mismatch when we
tried to query for the files with this path. Using the decomposed canonical equivalent will return a normalized String
after Unicode Normalization Form D which ensures that we have an identical String no matter the composition.

This solves crashes on concurrent parsing and thumbnailing situations, repeated thumbnailing of the same file, etc.

This resolves issues with parsing the same file multiple times, failing to trigger thumbnailing, etc.

(this simplifies the change of group identifiers and allows change even after compilation)

(cherry picked from commit 829f3c17)

(cherry picked from commit febbd62e)

(cherry picked from commit 447f2257)

(cherry picked from commit 57b825c6)