fix question answering error when running pipeline model on cuda

[imatiach-msft]

Description

fix question answering error when running pipeline model on cuda
Currently, QA pipelines passed to RAI text dashboard with cuda/gpu device fail in the explanations code. This PR allows question answering pipelines to be run on cuda in the RAI text dashboard, specifically when computing explanations. Also, passed device for vision package to explainer manager (instead of using default detected device).

Checklist

• I have added screenshots above for all UI changes.
• I have added e2e tests for all UI changes.
• Documentation was updated if it was needed.

[codecov-commenter]

Codecov Report

Merging #2291 (653fb06) into main (6edbfd3) will decrease coverage by 4.97%.
The diff coverage is 100.00%.

❗ Current head 653fb06 differs from pull request most recent head e4ba134. Consider uploading reports for the commit e4ba134 to get more accurate results

@@            Coverage Diff             @@
##             main    #2291      +/-   ##
==========================================
- Coverage   88.96%   84.00%   -4.97%     
==========================================
  Files         133       12     -121     
  Lines        7549      950    -6599     
==========================================
- Hits         6716      798    -5918     
+ Misses        833      152     -681     

Flag	Coverage Δ
unittests	84.00% <100.00%> (-4.97%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files Changed	Coverage Δ
...ext/responsibleai_text/utils/question_answering.py	93.54% <100.00%> (+0.21%)	⬆️

... and 121 files with indirect coverage changes