bench(evaluation): Re-introduce PERFLOG-bench for evaluation

In 87b252d, an evaluation benchmark was deleted meaning that the
PERFLOG.md could no longer track the performance of batch-evaluation of
a fixed-size data set. This reintroduces that benchmark.

The `par_batch_evaluate` is ~40 % faster than the previous instance, as
evidenced by 0.42.0-alpha.2's PERFLOG.md entry:

then:
Various Evaluations in 2^14 Points/Faster of the two/14
                        time:   [8.9896 ms 9.0391 ms 9.0978 ms]
                        thrpt:  [1.8009 Melem/s 1.8126 Melem/s 1.8225 Melem/s]

now:
Evaluation of degree-16383 polynomial in 16384 Points/Par batch-evaluate/14
                        time:   [6.4467 ms 6.6787 ms 6.9137 ms]

Author: Sword-Smith

twenty-first/benches/evaluation.rs

@@ -11,7 +11,8 @@ criterion_main!(benches);
 criterion_group!(
     name = benches;
     config = Criterion::default().sample_size(10);
-    targets = evaluation<{ 1 << 16 }, { 1 << 6 }>,
+    targets = evaluation<{ 1 << 14 }, { 1 << 14 }>,
+              evaluation<{ 1 << 16 }, { 1 << 6 }>,
               evaluation<{ 1 << 16 }, { 1 << 8 }>,
               evaluation<{ 1 << 19 }, { 1 << 6 }>,
               evaluation<{ 1 << 19 }, { 1 << 8 }>,
@@ -41,5 +42,8 @@ fn evaluation<const SIZE: usize, const NUM_POINTS: usize>(c: &mut Criterion) {
     let id = BenchmarkId::new("Entrypoint", log2_of_size);
     group.bench_function(id, |b| b.iter(|| poly.batch_evaluate(&eval_points)));
 
+    let id = BenchmarkId::new("Par batch-evaluate", log2_of_size);
+    group.bench_function(id, |b| b.iter(|| poly.par_batch_evaluate(&eval_points)));
+
     group.finish();
 }