data race fix

[OleksiiiSemko]

Fix of the #28 issue.
The Callstack structure is reset inside the CaptureContext destructor and deleted inside the unmap method. CaptureContext destructor and unmap method can be called in separate threads but std::unordered_map restore method is not threadsafe and the call is not protected by the critical section, because of this the Callstack destructor can be called twice for the same object.

[Neustradamus]

@OleksiiiSemko, @mattdurak: Any progress on it?

[mattdurak]

Seems safe, likely we would need to validate it for our use case, FYI @dcristoloveanu

[dcristoloveanu]

[dcristoloveanu]

It turns out that I just spend some time debugging exactly the same race.
My fix for it was to pass the callstack into map and remap rather than extract as an out arg pBlockInfo, but technically both fixes achieve the exact same thing, so I will take the fix from you @OleksiiiSemko.
For reference, the fix I did is in the crash_workaround branch.