refactor: fix undefined behavior and cleanup lock debugging code

[patricklodder]

This was triggered by and fixes #3305 (comment)

Removes undefined behavior by refactoring the way we destruct objects that track mutexes when debugging.

1. Prevent UB from LockData custom destructor by making it implicit
2. Replace the thread_specific_ptr holding the lock stack with a static map that stores it per thread.
3. Clarify types by introducing new aliases making the code more readable
4. Remove boost dependencies in favor of standard c++ for-loops, improving readability.

Backported from:

• 458992b
• 8d8921a
• f511f61
• 58e6881
• 26c093a

Note to testers: this functionality is only compiled in when using -DDEBUG_LOCKORDER, this is set with ./configure --enable-debug.

[chromatic]

All tests pass on x86-64 Linux and my first skim of the code looks correct. I'll review the linked commits in more detail.

[xanimo]

yeah no complaints re headless unit testing/qa on m2 👍

[xanimo]

i did however fail a couple times in a for range 10 loop for qa on m2 but i think we'll be able to iron out that stuff now that debugging on dogecoin binaries is upgraded + the getmocktime stuff i'm looking into 👍

[patricklodder]

I'm having some issues reproducing deadlock detection on macOS w/ a set of wip code that I know to have the issue (and does trigger both before and after this PR on x86_64 linux w/ gcc12). Because the code in question is networking related, and of course we use a newer boost on macOS arm64 local builds, and we use clang instead of gcc... there can be a handful of potential causes.

Therefore I'm putting this back into draft, pending addition of a unit test. I'm sorry for opening this too soon.

[patricklodder]

I've rebased this on top of 1.15.0-dev

[patricklodder]

I cherry-picked the unit test from 41b88e9, this is now ready for review.

I'm aware that there are multiple renames back and forth on this code that can influence picks; we can analyze these after this fix (as some of the picks in #3520 would make the UB problem come back; let's not do that)

[xanimo]

All unit and qa tests passed on x86_64-linux jammy. All unit tests passed on M2 but I encountered an issue with asyncore deprecation that I have to figure out. I don't think that issue should block this though.

[patricklodder]

asyncore depreciation

Yes, we'll have to fix that separately - not related to this code. Would you mind filing an issue with a log?

[xanimo]

asyncore depreciation

Yes, we'll have to fix that separately - not related to this code. Would you mind filing an issue with a log?

My pleasure.

[xanimo]

ACK. #3524