Add eval scripts for Clevr and Qlevr benchmarks

[jonathanlimsc]

1. Added eval scripts for calculating prediction accuracy for Clevr and Qlevr benchmarks

   • robin/eval/eval_vqa_clevr.py
   • robin/eval/eval_vqa_qlevr.py

2. Modified the inference scripts for Clevr and Qlevr to support question sampling during eval with a given seed

   • robin/eval/model_vqa_clevr.py
   • robin/eval/model_vqa_qlevr.py