Add screen functionality

[Jnelen]

This PR introduces a new screen function, enabling the ability to test multiple ligands (e.g., from a compound library) against a single target protein.

Key Features:

• Input Support:
  • Accepts protein input as a .fasta file (recommended) or a .pdb file.
  • Handles ligands from a directory of .sdf files, a single .sdf file, or .smi files containing ligand IDs and SMILES strings.
• Automated MSA Handling:
  • Automatically performs multiple sequence alignment (MSA) if a precomputed .m3a file is not provided, using the --use_msa_server option.
• Query Preparation:
  • Prepares a structured query directory automatically, making the process streamlined.
• Integration with predict:
  • Fully utilizes the existing predict functionality and supports all its regular features.

Example:

• To screen a .fasta file with a directory of .sdf files:
  boltz screen --protein protein.fasta --ligands ligand_directory --use_msa_server

This functionality is great for workflows involving large-scale ligand screening, as it simplifies things by automating MSA generation (when needed) and organizing input queries for you. I originally put this together for a virtual screening workflow and hope it can be useful to others too.

As for future improvements, it could be interesting to explore adding support for Singularity or SLURM for easier use on HPC systems. That might be better suited for a separate project, but I think this screen functionality fits nicely into the main repo and would be a solid addition.