Safe Parallel STL

[ashvardanian]

There are several implementations of the Parallel Standard Template Library (STL). The most widely adopted one, used by GCC, is based on Intel's oneAPI. Other modern alternative, such as NVIDIA's Senders, focus more on asynchrony. Every library has different goals. Fork Union focuses on:

• Exception-free noexcept operations
• No std::bad_alloc and hidden memory allocations
• Lock-free algorithms built on std::atomic
• Synchronous fork-join style parallelism

Given that Fork Union already provides primitives like for_n and for_n_dynamic, this PR poses a new direction:
Can we identify a core set of "standard-like" parallel algorithms that can be efficiently implemented the same way?

At this stage, this PR serves as a placeholder for that exploration.