Accelerated ts.optimize by batching Frechet Cell Filter

[falletta]

Changes

• Parallelized matrix exponential derivative (expm_frechet): Now processes batches of matrices simultaneously instead of looping through each one individually.
• Parallelized matrix logarithm (matrix_log_33): Now computes logarithms for all systems at once, grouping them by mathematical case type for efficient batch processing.
• Removed nested loops in cell force computation: The Frechet cell filter previously used two nested loops (one over systems, one over 9 directions). These are now replaced with a single batched operation that computes everything in parallel
• Added helper utilities: Small utility functions to handle the conversion between single-matrix and batched-matrix formats

The figure below shows the speedup achieved for 10-step atomic relaxation. The test is performed for a 8-atom cubic supercell of MgO using the mace-mpa model. Prior results are shown in blue, while new results are shown in red. The speedup is calculated as speedup (%) = (baseline_time / current_time − 1) × 100. We observe a speedup up to 564% for large batches.

[orionarcher]

Could we get some tests verifying identical numerical behavior of the old and new versions? Can be deleted before merging when we get rid of the unbatched version.