Add support for resuming interrupted simulations in optimize.py using BioDynaMo's backup/restore

[salva24]

Problem

Currently, optimize.py runs each Optuna trial via run_ABM(), which calls bdm run as a subprocess for the full simulated duration (total_minutes_to_simulate).

If a trial is interrupted partway through (e.g., killed by a cluster job scheduler or timeout), the entire simulation must restart from scratch. This wastes compute time and makes long calibration runs fragile.

Existing functionality

BioDynaMo already provides a native checkpoint/restore mechanism based on ROOT object serialization (WriteObject<bdm::Simulation>), exposed through:

• --backup <FILE> — periodically writes a full simulation snapshot
• --restore <FILE> — restores the simulation from a snapshot and continues execution

Proposed solution

Modify run_ABM() in optimize.py to:

• Pass a backup_file (and optionally backup_interval) as part of the generated params.json/configuration for each trial.
• Before launching bdm run, check whether a backup already exists for that trial instead of starting a new simulation.
• Remove the backup file after the trial finishes successfully to avoid accumulating stale checkpoints.

Benefits

• Allows long-running calibration trials to survive interruptions without losing progress.