Speed up packet dedup and fix benches

[jstarry]

Problems

• Packet deduping is slow for larger packets but we always add the full 1232 data buffer to the bloom filter
• It's slow to iterate over the bloom filter keys twice (once for contains and once for add)

Summary of Changes

• Only add the used data buffer portion of a packet to the bloom filter
• Modify AtomicBloom:add to return whether the item was newly added or not

Before

test bench_dedup_diff_big_packets   ... bench:  20,010,316 ns/iter (+/- 1,623,356)
test bench_dedup_diff_small_packets ... bench:  19,772,433 ns/iter (+/- 1,470,336)
test bench_dedup_same_big_packets   ... bench:  17,869,979 ns/iter (+/- 1,158,488)
test bench_dedup_same_small_packets ... bench:  18,181,612 ns/iter (+/- 2,449,875)

After

test bench_dedup_diff_big_packets   ... bench:  15,266,433 ns/iter (+/- 1,096,914)
test bench_dedup_diff_small_packets ... bench:   1,943,839 ns/iter (+/- 239,305)
test bench_dedup_same_big_packets   ... bench:  16,299,237 ns/iter (+/- 570,619)
test bench_dedup_same_small_packets ... bench:   3,363,462 ns/iter (+/- 2,007,400)

Fixes #

[sakridge]

lgtm

[mergify]

automerge label removed due to a CI failure

[codecov]

Codecov Report

Merging #22592 (4460c2c) into master (6edeed8) will increase coverage by 0.4%.
The diff coverage is 91.7%.

@@            Coverage Diff            @@
##           master   #22592     +/-   ##
=========================================
+ Coverage    81.1%    81.5%   +0.4%     
=========================================
  Files         560      555      -5     
  Lines      151206   149654   -1552     
=========================================
- Hits       122633   122086    -547     
+ Misses      28573    27568   -1005     

[t-nelson]

nice!