Fixed CLI errors.

[ShigrafS]

Closes #264

Flexible and Canonicalized Column Handling for dedup and sort

Overview

This pull request enhances the flexibility and robustness of column handling in pairtools, with a primary focus on improving CLI usability, internal consistency, and resilience to variations in input column names.

Note: This PR also fixes some flake8 linting issues.

Key Enhancements

✅ 1. Unified Column Lookup via headerops

• Introduced headerops.get_column_index() to allow CLI options like --c1, --c2, --p1, --p2, --s1, --s2, and --pt to accept both:
  • Integer indices (e.g., 1, 3) — with bounds and type checks.
  • String names (e.g., "chr1", "chrom1") — supporting canonicalization and case-insensitivity.
• Introduced headerops.canonicalize_columns() to standardize commonly used aliases (e.g., chr1 → chrom1, pt → pair_type) across all CLI tools and internal logic.

✅ 2. Improved dedup and sort CLI Behavior

• Replaced static string-based column name assumptions with dynamic lookups via get_column_index().
• Enabled seamless handling of extra_col_pair and extra_col options with warnings for missing columns instead of hard failures.
• Made the --pt (pair_type) option optional in sort, skipping it gracefully when not present in the header.
• Fixed incorrect help text for --c2 in dedup (was "Chrom 1 column", now corrected to "Chrom 2 column").

✅ 3. Column Defaults Remain String-Based

• CLI defaults (--c1, --c2, etc.) are still defined using canonical string names (e.g., "chr1", "pos1"), not integer indices as initially planned.
• However, the backend now supports either form due to get_column_index()'s flexibility.
• Suggested follow-up improvement: update default values to integers to fully align with the original plan.

✅ 4. Code Cleanup and Readability

• Replaced unclear variable names (e.g., l → line) for better readability across modules.
• Removed unused imports and deprecated warnings (e.g., cython backend placeholder).
• Refactored import ordering and string formatting for consistency.

✅ 5. Comprehensive Testing

• Added unit tests in test_headerops.py to validate:
  • Canonicalization of various column aliases.
  • Accurate index lookup from mixed input types (ints, strings, canonicalized names).
  • Proper error handling and edge case coverage (e.g., negative indices, invalid columns).
  • Integration with header extraction utilities.

---

Summary

This PR lays the groundwork for robust and user-friendly CLI interactions in pairtools, reducing the brittleness of column name handling and allowing greater flexibility for users working with varied input formats. It introduces modular utilities (canonicalize_columns, get_column_index) that can be reused across future tools and extensions.

---

Follow-Up Considerations

• [ ] Update CLI defaults (--c1, --c2, etc.) to use integer indices as per the original plan.

[ShigrafS]

@agalitsyna I've added a warning when sorting by columns not defined in pairsam_format.DTYPES_PAIRSAM, defaulting them to string type, and updated the --extra-col help text to reflect this behavior. I've also added a test to verify warnings for undefined custom columns
The PR is ready to be merged.
Kindly review it.