Xgboost bug fix

[cying111]

This PR is created to resolve the issues related to xgboost updates, specifically:

• update all label features to be type logical in both testSpliceSite and getTranscriptScores function
• update pre-saved models for both splicing junction prediction and defaultModels for read class scoring
• update test code for xgboost as well as saved outputs relevant to xgboost testing

[SuiYue-2308]

Hi, I’ve reviewed the code changes and everything looks good. I also tested it on a real dataset (dRNA), and it ran smoothly.

Update about xgboost pull request.

• In annotation-free mode, the se objects produced by the two branches are identical.
• Annotation-provided mode, the se objects are no longer identical.
  (1) NDR score is similar (0.04597701 v.s. 0.04595099)
  (2) Novel transcripts: The number of novel transcripts differs between XGBoost versions (340 with XGBoost 1.7.11.1 vs. 345 with XGBoost 3.1.3.1). Among these, 279 transcripts have identical rowRanges.
  (3) Expression levels: Transcript-level read counts are largely similar; the differences might from XGBoost update.

[jonathangoeke]

please add release note. Thanks!