tests: compute score for gapfills
Currently, at the end of a gapfill test (e.g., https://languagelab.be/tests/4 ), the score shown is 100%.
We could compute the score per correct letter (e.g., beauti, response "fol" = 2/3 = .66 for this item) or per gap (response "fol" is incorrect or incomplete so = 0/3 = 0 for this item -> binary correct/incorrect for this gap). The second option (binary) is probably sounder.