bug fix: string truncation in AtomArray annotations

[davegrays]

The structure.array() method truncates any string-based annotations to an arbitrary length of string. i.e. you can end up with

atoms[1000] --> Atom(..., str_annot="111", ...)
atoms[1000].str_annot.dtype --> dtype('<U3')

array = structure.array(atoms)
array[1000] --> Atom(..., str_annot="1", ...)
array[1000].str_annot.dtype --> dtype('<U1')

The culprit is this
array.add_annotation(name, dtype=type(atoms[0]._annot[name]))
is using the first atom in the array, which is often the shortest, to determine the dtype of the whole array. This is particularly problematic when processing cif columns that contains string representations of ids. Instead, we need to find the maximum length across the atom list to ensure no truncations occur.

[codspeed-hq]

CodSpeed Performance Report

Merging #755 will not alter performance

_{Comparing davegrays:fix/atomarray-string-annotations (1effc13) with main (d7df095)}

Summary

✅ 59 untouched benchmarks

[padix-key]

Thanks for the fix 👍!

[padix-key]

The currently failing tests are caused be a change on the side on the RCSB, which will be addressed in #759. Hence, I will merge this PR.