Fix PDB download

[linusyh]

Issue:

• workshop download pdb gets stuck, failing to terminate.

Cause:

• The pretein pdb_id: 8CKB is not available in the mmtf format, as verified with the official mmtf-python API. This causes rcsb.fetch to throw RequestError which, when uncaught, causes the entire download process to hang.

Solution:

• Catch RequestError in download_pdb_mmtf
• Attempt retries in case of network issues

During implementation, I found that download_pdb_mmtf is duplicated in datasets/utils and scripts/download_pdb_mmtf.

• Remove code duplication

[chaitjo]

That's a good catch that this particular protein is not downloadable. The last time we'd downloaded the PDB, this entry was not deposited so the error was not caught.

@a-r-j, @amorehead, is there a way to set a cutoff date later than which we do NOT download any newer PDBs via PDB manager?

[a-r-j]

Thanks for the fix @linusyh

@chaitjo The download of the PDB is intended to function as a local copy so the correct behaviour would be to download any entries that are downloadable. Examples in downstream task datasets won't drift & this should automatically handle PDB code deprecation as structures are replaced with newer entries.

[a-r-j]

@linusyh could you add a note to the changelog please