[src/snmalloc/ds_core/redblacktree.h] Make free-standing

[SamuelMarks]

Closes #670.

Taking a quick look:

$ rg -IFNth --trim '#include <' | sort -u
#include <array>
#include <atomic>
#include <chrono>
#include <climits>
#include <cstddef>
#include <cstdint>
#include <cstdlib>
#include <cstring>
#include <errno.h>
#include <fcntl.h>
#include <functional>
#include <iomanip>
#include <iostream>
#include <limits>
#include <new>
#include <sstream>
#include <stddef.h>
#include <stdio.h>
#include <string>
#include <string.h>
#include <strings.h>
#include <string_view>
#include <sys/mman.h>
#include <sys/uio.h>
#include <type_traits>
#include <unistd.h>
#include <utility>

I'm not sure how much of this can obviously be removed without looking really wonky in the C++ world.

Maintainers / @SchrodingerZhu - do you just want a C port of this?

[nwf]

I think I'd prefer that the debugging code in redblacktree.h, which seems to be the sole consumer of <string>, as well as its #include, be made conditional on __STDC_HOSTED__ being nonzero. (See https://en.cppreference.com/w/cpp/freestanding). @mjp41?

ETA: We should also have at least one -ffreestanding CI build, if that's something people care about. :)

[mjp41]

In this case this is just code that is executed by the test harness to unit test the RB tree. I would sooner this be under and #ifdef, so that it isn't compiled with the library just the unit tests.

[mjp41]

I have redone your idea, but with stack allocation in #680, and also copied the original contents into the string.

I would sooner not take a dependence on malloc/free in this code. The std::string did have a dependence on an allocator, which is not nice for allocator debug code.