Introducing a C++ Backend for OCaml Compilation

I've developed a new C++ backend for ocamlc, enhancing the current C backend used by the runtime and FFI. This update allows OCaml programs to be compiled into idiomatic C++ code. For instance, a program calculating prime numbers can now be translated into C++ using the command `ocamlc -incr-c primes.ml`, resulting in a readable `primes.cpp`. This C++ code, however, requires a C++ interpreter like g++ to execute, and it supports passing arguments to the main function using the -D option.

C++'s purely functional nature means the OCaml standard library, which uses mutable state, is unavailable. Therefore, I reimplemented parts of the List module in a purely functional style to overcome this limitation. Running the program with `g++ -Dlimit=100 primes.cpp` should print prime numbers below 100, though the output format may seem unusual due to C++'s historical roots as a C preprocessor.

C++ does not support OCaml's infix :: syntax for list cons cells, so the output must be read as nested Cons<hd, tl> cells. Larger computations can be challenging for C++, but enabling the -ftemplate-depth=999999 option allows for more extensive programs. On my machine, this setup prints primes up to 10000 in about 30 seconds, using 11 GiB of memory.

Performance varies across C++ implementations. For example, clang++ is more efficient than g++, completing the task in a second with minimal memory usage. However, the primary performance issue is algorithmic. A more efficient algorithm, based on O'Neill's priority-queue method and Okasaki's leftist heap, significantly reduces computation time and memory usage. With these improvements, g++ computes primes below 10000 in just 8 seconds, using only 3.1 GiB of memory.

Looking ahead, this approach could extend to other languages. Once Rust supports partial impl specialization, it might also run OCaml programs.