Gerbil QA swaps annotator and dataset in JSON LD output

# Problem

In its JSON LD, Gerbil QA seems to confuse the annotator and dataset if both are uploaded files.

```
{
"@graph" : [ {
"@id" : "http://gerbil-qa.aksw.org/gerbil/experiment?id=202604100023",
"@type" : [ "gerbil:Experiment", "qb:Dataset" ],
"structure" : "gerbil:dsd",
"experimentType" : "gerbil:QA",
"matching" : "gerbil:StrongEntityMatch",
"label" : "Experiment 202604100023"
}, {
"@id" : "http://gerbil-qa.aksw.org/gerbil/experimentTask_409109",
"@type" : "qb:Observation",
"dataset" : "http://gerbil-qa.aksw.org/gerbil/experiment?id=202604100023",
"Macro_F1_QALD" : "1.0",
"annotator" : "http://gerbil-qa.aksw.org/gerbil/dataId/annotators/gold_1.json_(uploaded)",
"doc" : "0.0",
"gerbil:dataset" : "http://gerbil-qa.aksw.org/gerbil/dataId/corpora/pred_file_(uploaded)",
"errorCount" : "0",
"language" : "http://gerbil-qa.aksw.org/gerbil/dataId/languages/en",
"macroF1" : "1.0",
"macroPrecision" : "1.0",
"macroRecall" : "1.0",
"microF1" : "1.0",
"microPrecision" : "1.0",
"microRecall" : "1.0",
"statusCode" : "0",
"timestamp" : "2026-04-10T21:53:00.369Z"
},
```

# Expected behavior

The IRIs should be used for the correct predicates.