Fix for concurrency bug

[miclegr]

Description

When adding a point to an index without providing an id, we rely on TypedIndex::currentLabel to get one. But we are not doing it in a threadsafe way. This PR fixes that, solves #65

Changes Made

C++

TypedIndex::currentLabel is now std::atomic and it's post-incremented atomically to get a new id when needed

Testing

Checklist

• My code follows the code style of this project.
• I have added and/or updated appropriate documentation (if applicable).
• All new and existing tests pass locally with these changes.
• I have run static code analysis (if available) and resolved any issues.
• I have considered backward compatibility (if applicable).
• I have confirmed that this PR does not introduce any security vulnerabilities.

Additional Comments

[miclegr]

C++ linting failing due to this

[markkohdev]

Left a question about the usage of the atomic. I'm not sure this will result in behavioral parity with the existing implementation

[markkohdev]

Is .fetch_add(1) actually what we want throughout this change? Wouldn't that result in us incrementing currentLabel multiple times where we actually only wanted to increment it once?

[miclegr]

Correct, I'm reverting this

[miclegr]

Actually not, my original implementation is correct. Basically here we assign an id, then here:

voyager/cpp/TypedIndex.h

Line 384 in 33d68cd

start = 1;

we set start = 1 so that when we loop via ParallelFor we start from 1:

voyager/cpp/TypedIndex.h

Line 392 in 33d68cd

ParallelFor(start, rows, numThreads, [&](size_t row, size_t threadId) {

therefore we need currentLabel already incremented

[markkohdev]

👍 Thanks for addressing the comment! This looks good to me, albeit race conditions are always a little tricky to verify fixes for