Optimise particle I/O with chunking and add verbosity control

[WillJRoper]

Implements chunked particle reading to minimize I/O operations and adds configurable output verbosity. More specifically:

1. Chunked Particle Reading

Optimised particle I/O by reading contiguous chunks instead of individual cells:

• Added ParticleChunk struct to track contiguous cell ranges
• Implemented gap-filling strategy: merge nearby chunks if gap < gap_fill_fraction * total_particles
• New gap_fill_fraction parameter controls tradeoff between unused particle reads and I/O call count
• Reduces HDF5 read operations by orders of magnitude for sparse grids

Performance impact: Fewer I/O calls at the cost of reading some unused particles (controlled by gap fraction).

2. Verbosity Control

Added optional verbosity argument for cleaner output:

• 0 = errors only
• 1 = rank 0 only (default)
• 2 = all ranks

Usage: ./parent_gridder params.yml 8 42 1

3. Code Cleanup

• Removed unnecessary grid point percolation code (grid points never leave top-level cells)
• Simplified output functions by ~50 lines
• Added debug checks for grid point distribution invariants