playground: speed up pdf loading

[miguelgrc]

Closes cern-sis/issues-inspire#977

• Now fetching PDFs from inspire when available, otherwise falling back to arxiv
• PDFs for all papers are fetched immediately and in parallel when the LLM search response is received, and they are cached locally as blobs. While loading, they will be saved as pending.
• When opening a paper, the PDF viewer will load all the PDFs from the cache in parallel into multiple pdf viewers, only one of them being visible (this is to avoid pdf rendering times when switching back and forth between papers). If any PDF is pending, it will just await it to avoid duplicate requests, otherwise (if for some reason it hasn't been fetched at all) it will retry, and if it doesn't work it will display an error.

With these changes PDFs load faster and switching back and forth between papers is almost instant.

[PascalEgn]

Looks good! just one small question.