Fix Multi-Threaded Physics 2D crashes

[Listwon]

Fixes #29369, prevents crash in #29364 (but there are many other errors in log), possible fix for #35017 and #47489

I know I didn't add this to master first, but all test MRPs are for previous versions and privately I need this for 3.x first too.

I need some guidance how to rework this fix - I'm not sure if it's better to create separate data structures (as I did in map_mt.h) and replace them in critical places (area_2d_sw.h, area_2d_sw.cpp) or directly change existing structures as I did in self_list.h (maybe with addition of some directives or template arguments).

Also I'll appreciate any help testing this.

[pouleyKetchoupp]

Thanks for working on this! Your contribution is welcome, but I wonder if this is the right approach in fixing these threading issues.

Multi-threading issues can be complex so team work is going to be important here :)

Thread-safe data structures:
Adding mutexes on single operations within map or self_list is not the most efficient way to solve multi-threading issues. A better approach could be to add a mutex in the code that uses the structure, so it can lock only once when needed and avoid to possibly stall multiple times.

About the specific issues in physics:
Maybe a mutex will be needed in some cases, but I'd like to check if simpler (and more efficient for performance) strategies can be used:

• As mentioned in Crash in destructor AreaPair2DSW with Multi-Threaded physics #29369 (comment), some of these crashes can be solved by avoiding certain calls in _process and doing them in _physics_process instead. Maybe we could have errors in some calls, based on Engine::is_in_physics_frame() (possibly only when physics is running on a separate thread) to prevent cases that would crash.
• I wonder if Make Servers truly Thread Safe #45852 fixes some of the crashes on master

So I think the next steps can be to test these issues on master and identify the ones that are still problematic, so we can decide how to solve these cases individually.

About testing:
What have you been able to test/repro so far? Apart from testing the MRP from different issues, it seems you can compile godot with use_tsan=yes in order to enable thread sanitizer and catch specific threading issues (like the case from #47489). I haven't used it before, but maybe @qarmin can help if you have more questions.
I'll be happy to help testing once I have this set up as well.

[pouleyKetchoupp]

Closing this PR as there's no update and it's not likely to be merged the way it's implemented.