Add lightweight PDB (Protein Data Bank) file support

[behroozazarkhalili]

Summary

This PR adds support for loading PDB (Protein Data Bank) files directly with load_dataset().

PDB is the legacy fixed-width format for representing 3D macromolecular structures, widely used for historical datasets and still common in computational biology workflows.

Key Features

• Zero external dependencies - Pure Python parser using fixed-width column positions per official PDB specification
• Record type filtering - Load ATOM, HETATM, or both record types
• Column selection - Choose specific columns to reduce memory usage
• Compression support - Automatic detection of gzip, bzip2, and xz compressed files via magic bytes
• Batch processing - Configurable batch size for memory-efficient processing

Columns (ATOM/HETATM records)

Column	Type	Description
record_type	string	ATOM or HETATM
atom_serial	int32	Atom serial number
atom_name	string	Atom name (e.g., CA, N, C)
residue_name	string	Residue name (e.g., ALA, GLY)
chain_id	string	Chain identifier
residue_seq	int32	Residue sequence number
x, y, z	float32	Coordinates (Å)
occupancy	float32	Occupancy factor
temp_factor	float32	Temperature factor (B-factor)
element	string	Element symbol

Supported Extensions

.pdb, .ent (and compressed variants)

Usage

from datasets import load_dataset

# Load PDB file
dataset = load_dataset("pdb", data_files="structure.pdb")

# Load only ATOM records (exclude ligands/water)
dataset = load_dataset("pdb", data_files="structure.pdb", record_types=["ATOM"])

# Load specific columns
dataset = load_dataset("pdb", data_files="structure.pdb", 
                       columns=["atom_name", "residue_name", "x", "y", "z"])

Use Cases

• Legacy structure dataset processing
• Molecular dynamics trajectory analysis
• Structure-based ML training data
• Protein visualization data preparation

References

• PDB format specification: https://www.wwpdb.org/documentation/file-format

Test Results

All 24 tests pass:

• Basic loading, column filtering, record type filtering
• Gzip compression, multi-chain structures, alternate locations
• Charged atoms, batch sizes, schema types, feature casting
• Empty files, multiple files, insertion codes, negative coordinates

Part of the bioinformatics file format support series (FASTA #7923, FASTQ #7924, mmCIF #7925).

cc @georgia-hf