perf(codspeed): Add benchmark suite with CI workflow

[cbueth]

Add benchmarks using CodSpeed. Cover the core connectivity metrics and network reconstruction. It uses the simulation mode for consistent, comparable perf numbers.

• 18 reconstruct_network benchmarks (lc/rc/gc/gv across 10, 20, 40, 50 nodes + parallel workers)
• 6 isolated metric benchmarks (lc/rc/gc/cop/mi/te)
• 1 find_optimal_lag bench
• CI workflow that runs them on every push/PR via CodSpeedHQ/action@v4
• pytest-codspeed in test deps

Benchmarks auto-skip unless --codspeed is passed, so uv run pytest stays fast for normal dev.

[codspeed-hq]

Congrats! CodSpeed is installed 🎉

🆕 17 new benchmarks were detected.

You will start to see performance impacts in the reports once the benchmarks are run from your default branch.

Detected benchmarks

• test_continuous_metric[lc]: 2.2 ms
• test_continuous_metric[rc]: 2.5 ms
• test_discrete_metric[mi]: 9 ms
• test_discrete_metric[te]: 35.1 ms
• test_reconstruct[nodes5-gc]: 67.6 ms
• test_reconstruct[nodes5-gv]: 67.6 ms
• test_reconstruct[nodes5-lc]: 38.4 ms
• test_reconstruct[nodes5-rc]: 44.2 ms
• test_find_optimal_lag: 143.6 µs
• test_granger_f_test: 3.3 ms
• test_ordinal_patterns: 13.4 ms
• test_reconstruct[nodes10-gc]: 300.7 ms
• test_reconstruct[nodes10-gv]: 309.6 ms
• test_reconstruct[nodes10-lc]: 170.7 ms
• test_reconstruct[nodes10-rc]: 198.6 ms
• test_reconstruct_workers[par2]: 7.2 ms
• test_reconstruct_workers[seq]: 38.3 ms

---