backport: bitcoin#20942 (Move some net_processing globals into PeerManagerImpl)

[kwvg]

Split off due to TSan failures and depends on #5352

[github-actions]

This pull request has conflicts, please rebase.

[PastaPastaPasta]

Marking as ready so hopefully CI runs..?

[PastaPastaPasta]

CI just doesn't want to run here... @kittywhiskers can you open a new pr from this commit? thanks

[UdjinM6]

"removing assignee" + "rebasing" + "dropping the branch in gitlalb repo" + "webhook redelivery" helped it seems. easy 😅

[knst]

The difference that I noticed between this PR and original bitcoin's changes.
All mentioned points are checked and valid, just a notice for history/re-review if needed.

Checked and correct as expected - just moved to header:

• ProcessGetData - missing EXCLUSIVE_LOCKS_REQUIRED(!cs_main)
• PeerManagerImpl::MarkBlockAsReceived - missing EXCLUSIVE_LOCKS_REQUIRED(cs_main)
• PeerManagerImpl::MarkBlockAsInFlight - missing EXCLUSIVE_LOCKS_REQUIRED(cs_main)
• PeerManagerImpl::MaybeSetPeerAsAnnouncingHeaderAndIDs - missing EXCLUSIVE_LOCKS_REQUIRED(cs_main)
• PeerManagerImpl::TipMayBeStale - missing EXCLUSIVE_LOCKS_REQUIRED(cs_main)
• PeerManagerImpl::FindNextBlocksToDownload - missing EXCLUSIVE_LOCKS_REQUIRED(cs_main)

Checked and correct as expected - it's not changed line in this PR (both bitcoin and our):

• m_connman.ForNode(nodeid, [this](CNode* pfrom) - - missing EXCLUSIVE_LOCKS_REQUIRED(::cs_main) BUT ASSERT for cs_main

Checked and correct as expected - it's appliable only for wtx

• missing LOCK(cs_main); before "ignoring duplicate wtxidrelay from peer=%d\n", pfrom.GetId())"

[UdjinM6]

these unrelated changes (some mentioned by @knst) must be reverted cd087fc imo

[kwvg]

I mean, redundant annotations were eventually removed with bitcoin#21188 and it is clear that its presence is more a liability than anything.

I removed it because my understanding is that the annotations need to be present in the declaration and having to change the annotation in the definition as well, while debugging, made it a bit of drag. So I opted to remove to as a matter of convenience. The removal in the bitcoin PR can be backported once its preceding contents are also backported but it's also done via a scripted-diff so nothing is lost in terms of ease-of-backporting if it is OOO.

[UdjinM6]

hmm, ok, makes sense. utACK

[PastaPastaPasta]

utACK for merging via merge commit;

all checks passing