Fix permutation column offsets and add coverage

[rjbohlender]

Summary

• fix the permutation slice logic so each run writes to a unique region in the shared matrix
• add a regression-style test that loads two runs and verifies permutations retain their columns
• provide a second sample run file used by the new test

Testing

• pytest
• python IBDReduce/ibdreduce_v3.py --prefix tests/data/sample_chr --suffix "{i}_run{j}.txt" --nperm 2 --nruns 2 --gmap tests/data/gmap --output after.txt --single 1
• python IBDReduce/ibdreduce_v3.py --prefix tests/data/sample_chr --suffix "{i}_run{j}.txt" --nperm 2 --nruns 2 --gmap tests/data/gmap --output after_repeat.txt --single 1

---

https://chatgpt.com/codex/tasks/task_e_68d59aae10648320964c4e4f0ce37ddf