web_service: fix covalent position validation

[dajiaohuang]

Summary

• validate polymer covalent-bond positions against the actual sequence length
• keep the existing positive-position check for every entity type
• defer non-polymer upper-bound validation to the inference parser, which has parsed ligand residue and atom information
• add a focused viewer regression test for a terminal protein residue bonded to a ligand

Root cause

ProtenixInputViewer.get_result() unwrapped each entity to its inner metadata dictionary and used len() on that dictionary. A polymer dictionary normally contains three fields, so valid sequence positions greater than 3 were rejected before featurization.

Testing

• python -m pytest tests/test_web_service_covalent_bonds.py -q (1 passed)
• python -m unittest tests.test_web_service_covalent_bonds -v (1 passed)
• python -m py_compile protenix/web_service/viewer.py tests/test_web_service_covalent_bonds.py
• python -m flake8 protenix/web_service/viewer.py tests/test_web_service_covalent_bonds.py --count --select=E9,F63,F7,F82 --show-source --statistics (0 errors)

Closes #341