Use faer backend and ensure Python deps for tests

[SauersML]

Summary

• switch the crate to prefer the faer backend and gate ndarray-linalg usage behind the OpenBLAS/MKL feature flags
• add a reusable Python dependency bootstrapper and invoke it from the PCA and EigenSNP integration tests
• rewrite the PCA tests to use Gram–Schmidt and faer eigenvalue helpers instead of ndarray-linalg utilities

Testing

• cargo test --no-default-features --features backend_faer (fails: large EigenSNP correlation/projection checks)

---

https://chatgpt.com/codex/tasks/task_e_68d331ecae80832eb9503d1de01e6edd