ed25519_verify_cpu: remove allocation in hot path

[alessandrod]

The old code was a remnant from a time when we tried to "balance" the amount of work across threads by re-batching and therefore allocating.

This is a hot path. Allocations are bad. Let work stealing work its magic.

before

running 8 tests
test bench_sigverify_high_packets_large_batch   ... bench:  16,990,766.60 ns/iter (+/- 3,061,142.38)
test bench_sigverify_high_packets_small_batch   ... bench:  17,250,462.50 ns/iter (+/- 2,758,672.55)
test bench_sigverify_low_packets_large_batch    ... bench:   9,067,963.60 ns/iter (+/- 136,488.70)
test bench_sigverify_low_packets_small_batch    ... bench:   9,054,538.40 ns/iter (+/- 125,983.52)
test bench_sigverify_medium_packets_large_batch ... bench:  17,483,717.50 ns/iter (+/- 6,094,133.22)
test bench_sigverify_medium_packets_small_batch ... bench:  18,705,141.90 ns/iter (+/- 5,872,418.63)
test bench_sigverify_simple                     ... bench:  20,167,362.70 ns/iter (+/- 2,705,726.58)
test bench_sigverify_uneven                     ... bench:   7,192,356.90 ns/iter (+/- 815,398.52)

after

running 8 tests
test bench_sigverify_high_packets_large_batch   ... bench:  10,494,283.30 ns/iter (+/- 852,231.77)
test bench_sigverify_high_packets_small_batch   ... bench:  10,249,499.70 ns/iter (+/- 226,444.85)
test bench_sigverify_low_packets_large_batch    ... bench:     465,018.57 ns/iter (+/- 21,994.14)
test bench_sigverify_low_packets_small_batch    ... bench:     461,375.30 ns/iter (+/- 24,855.80)
test bench_sigverify_medium_packets_large_batch ... bench:   2,650,445.62 ns/iter (+/- 177,836.54)
test bench_sigverify_medium_packets_small_batch ... bench:   2,779,869.52 ns/iter (+/- 145,892.43)
test bench_sigverify_simple                     ... bench:     768,623.86 ns/iter (+/- 56,839.41)
test bench_sigverify_uneven                     ... bench:   1,325,557.05 ns/iter (+/- 92,470.47)

[alessandrod]

on a single node loaded with bench-tps, almost doubles the amount of valid packets that make it past sigverify

[bw-solana]

The work stealing can happen within a packet batch?

[alessandrod]

The work stealing can happen within a packet batch?

Yes because we're flattening

[apfitzge]

The work stealing can happen within a packet batch?

Yes because we're flattening

Oh. Hmm I thought this was not true, but it is: https://docs.rs/rayon/latest/rayon/iter/trait.ParallelIterator.html#method.flatten

[apfitzge]

lgtm

[mergify]

Backports to the beta branch are to be avoided unless absolutely necessary for fixing bugs, security issues, and perf regressions. Changes intended for backport should be structured such that a minimum effective diff can be committed separately from any refactoring, plumbing, cleanup, etc that are not strictly necessary to achieve the goal. Any of the latter should go only into master and ride the normal stabilization schedule. Exceptions include CI/metrics changes, CLI improvements and documentation updates on a case by case basis.