Expose AllWave's full sparsification interface

[ekg]

Summary

Exposes all of AllWave's sparsification strategies through the CLI, following the AllWave naming model instead of our custom simplified interface.

New Sparsification Options

# All pairs (default)
--sparsify none

# Random sampling
--sparsify random:0.5          # 50% of pairs

# Connectivity-based (Erdős-Rényi)
--sparsify connectivity:0.95   # 95% probability of connected component

# Tree sampling (intelligent pair selection)
--sparsify tree:3,3,0.1,16     # k_nearest, k_farthest, random_fraction, kmer_size

Tree Sampling Format

tree:K,K2,F,SIZE where:

• K = k-nearest neighbors per sequence (1 = minimum spanning tree, higher = denser)
• K2 = k-farthest neighbors (strangers for diversity)
• F = additional random sampling fraction (0.0-1.0)
• SIZE = kmer size for Mash distance calculation (default: 16)

Example: tree:3,3,0.1,16 creates:

• 3 nearest neighbor edges per sequence
• 3 farthest neighbor edges per sequence
• 10% additional random pairs
• Uses 16-mer for distance calculation

Iterative Mode Integration

Iterative mode now respects the user's --sparsify setting:

# Use custom tree sampling for iterative alignment
./seqrush -s input.fa -o output.gfa --iterative --sparsify tree:5,5,0.2,16

If no tree sampling specified, iterative mode defaults to tree:3,3,0.1,16 and shows a note.

Changes

• Changed default from '1.0' to 'none' (more intuitive)
• Added comprehensive parsing for all AllWave strategies
• Updated help text with examples
• Iterative mode respects user's sparsification settings
• Backward compatible: plain floats treated as random factors (with deprecation warning)

Testing

✅ Build successful
✅ All tests pass
✅ Help text updated
✅ Iterative mode uses parsed settings