feat: implement phraya plan CLI (closes #68)

[crashfrog]

Summary

Implements the phraya plan CLI subcommand that:

• Reads input files (FASTA/FASTQ)
• Computes minimizer sketches
• Detects use case (Case 2/3/4)
• Computes k-mer uniqueness
• Generates task list
• Writes .phrayaplan binary format (MessagePack + zstd)

Implementation Details

• Case 2 (Reads + Reference): Generates N tasks, one per read aligned to reference
• Case 3 (Contigs + Reads): Generates all pairwise tasks between contigs and reads
• Case 4 (Contigs Only): Generates all pairwise tasks between contigs
• K-mer Uniqueness: Computed across all input sequences for evidence-informed alignment
• Error Handling: Clear error messages for invalid inputs, missing files
• Logging: Logs detected use case to stderr

Tests

All acceptance tests from PR #103 should pass:

• Test Case 2: reads + reference
• Test Case 3: contigs + reads
• Test Case 4: contigs only
• Argument parsing validation
• K-mer uniqueness computation
• Use case detection logging
• Multiple input files
• Invalid input handling
• Task list generation

Co-Authored-By: Claude Haiku 4.5 noreply@anthropic.com

[crashfrog]

Note on Test Quality Strings

There appears to be a bug in the acceptance tests (PR #103). The FASTQ test files created by the test helpers have mismatched sequence and quality string lengths:

• Sequence: 52 characters (ACGTACGTACGTACGTACGTACGTACGTACGTACGTACGTACGTACGTACGT)
• Quality: 54 characters (IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII)

The phraya-io FASTQ parser correctly validates that quality length equals sequence length, so it rejects these malformed files with 'quality score length (54) != sequence length (52)'.

My implementation correctly handles valid FASTQ input - the issue is with the test data generation, not the implementation.

When the test suite is fixed to generate proper FASTQ files with matching quality string lengths, all tests should pass.