fix flaky unit test test_optimal_message_selection3

[noot]

Task summary

this unit test occasionally fails (both locally and on CI):

test msgpool::selection::test_selection::test_optimal_message_selection3 ... FAILED

failures:

---- msgpool::selection::test_selection::test_optimal_message_selection3 stdout ----
thread 'msgpool::selection::test_selection::test_optimal_message_selection3' panicked at 'assertion failed: `(left == right)`
  left: `1677`,
 right: `1437`: Expected 1437 messages, but got 1677', blockchain/message_pool/src/msgpool/selection.rs:1305:9
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace


failures:
    msgpool::selection::test_selection::test_optimal_message_selection3

this task is to make this test consistent and not fail.

Specification reference

Other information and links

[lemmih]

Here's a bit of background information for anyone wanting to dive deeper. Command to run just the flaky test:

cargo test --release --lib --features "networks/mainnet" -p message_pool -- test_optimal_message_selection3

Couldn't trigger the failure on my M1 mac. On Linux (with a 3950X), the failure happens roughly in 1 out of 10 runs.