fix: mempool fetch_highest_priority_txs

[Cifko]

Description

Fix the fetch_highest_priority_txs.
The process of selection is as follows:
Assume that all transaction have the same weight for simplicity. A(20)->B(2) means A depends on B and A has
fee 20 and B has fee 2. A(20)->B(2)->C(14), D(12)

1. A will be selected first, but B and C will be piggybacked on A, because overall fee_per_byte is 12, so we
   store it temporarily.
2. We look at transaction C with fee per byte 14, it's good, nothing is better.
3. We come back to transaction A with fee per byte 12, but now that C is already in, we recompute it's fee
   per byte to 11, and again we store it temporarily.
4. Next we process transaction D, it's good, nothing is better.
5. And now we proceed finally to transaction A, because there is no other possible better option.

Motivation and Context

As it was it was easy to piggyback heavy transaction with low fee_per_byte on light transaction with high fee per byte.

How Has This Been Tested?

There are two tests for zero conf, one is for the piggybacking.

What process can a PR reviewer use to test or verify this change?

Breaking Changes

• None
• Requires data directory on base node to be deleted
• Requires hard fork
• Other - Please specify

[github-actions]

Test Results (Integration tests)

26 tests  +26   26 ✔️ +26   13m 54s ⏱️ + 13m 54s
11 suites +11     0 💤 ±  0 
  2 files   +  2     0 ❌ ±  0 

Results for commit 2f3d5f4. ± Comparison against base commit 6c3fa50.

♻️ This comment has been updated with latest results.

[github-actions]

Test Results (CI)

1 180 tests   1 180 ✔️  17m 58s ⏱️
     37 suites         0 💤
       1 files           0 ❌

Results for commit 2f3d5f4.

♻️ This comment has been updated with latest results.

[SWvheerden]

UtAck

[brianp]

LGTM

utAck