Optimize ByteTagList by tracking byte tags boundaries This patch restores performance to the level nearly equal to that before recent bugfixing, as was checked by running utils/bench-packets.

merge

run the benchmark over multiple iterations to minimize variance

Remove ByteTagList::IsDirty{Start,End} These functions are only used once and do not save time. Both functions iterate over the whole list just to prevent iterating over the whole list if they return false. As traversing the list is required in any case, these functions can be removed.

rescan network bindings

When data is written with WriteU64, it should be tested with ReadU64, not ReadLsbtohU64.

Make Buffer::CreateFullCopy private

Fragmentation and concatenation benchmark

Optimize Buffer::Iterator::CheckNoZero

remove un-needed return value