New throughput performance example

[arthurpaulino]

Add example to measure throughput in it/s (raw and adjusted) across different rcs and number of folding steps

This is a PoC that we might want to later integrate with criterion somehow.

On a short run with LURK_RC=8,16 LURK_N_FOLDS=2 cargo run --release --example tp_table, it prints:

Raw throughput
┌────────────┬─────────────┬────────────┬────────────┐
│ rc\n_folds │ 0           │ 1          │ 2          │
├────────────┼─────────────┼────────────┼────────────┤
│          8 │  85.61±3.15 │ 64.45±0.92 │ 56.22±0.68 │
│         16 │ 111.48±2.30 │ 77.47±0.70 │ 67.54±0.61 │
└────────────┴─────────────┴────────────┴────────────┘
Adjusted throughput
┌────────────┬────────────┬────────────┐
│ rc\n_folds │ 1          │ 2          │
├────────────┼────────────┼────────────┤
│          8 │ 51.72±1.19 │ 48.00±0.75 │
│         16 │ 59.38±0.83 │ 56.42±0.63 │
└────────────┴────────────┴────────────┘

Note: these numbers are from my weak machine.

[porcuquine]

Since we're already using this as an 'indicator' of something, I think it's fine to merge (though I would like to fix the 'source-code case' issue as a matter of good policy in examples) as-is. Before we start taking it too seriously, it would be useful to document the intended behavior and the implementation so:

• users know exactly what this purports to measure
• reviewers can verify that it actually does so

This review claims to have done neither of those but notes that as an n+1 'number go up' example, we're already getting some benefit out of running it from time to time.