64-byte alignment causes assertion in rbtree_best_fit

[tamboril]

Trying to make managed heap memory with a 64-byte alignment causes the assertion shown below the code.

This asserts when align > 32 in boost 1.66.0 on CentOS7-64/gcc 7.2:

#include <boost/interprocess/mem_algo/rbtree_best_fit.hpp>
#include <boost/interprocess/managed_heap_memory.hpp>

using namespace ::std;
using namespace boost::interprocess;

int main(int, char *argv[])
{
    constexpr const static size_t align = 64;

    typedef basic_managed_heap_memory<
        char,
        rbtree_best_fit<mutex_family, offset_ptr<void, long, unsigned long, align>, align>,
        iset_index
    > synchronized_managed_heap_t;

    synchronized_managed_heap_t mem(10000); // asserts when align > 32

    return 0;
}

boost/interprocess/mem_algo/rbtree_best_fit.hpp:410: void boost::interprocess::rbtree_best_fit<MutexFamily,
VoidMutex, MemAlignment>::priv_add_segment(void*, boost::interprocess::rbtree_best_fit<MutexFamily, VoidMutex, MemAlignment>::size_type)
 [with MutexFamily = boost::interprocess::mutex_family; VoidPointer = boost::interprocess::offset_ptr<void, long int, long unsigned int, 64>;
long unsigned int MemAlignment = 64; boost::interprocess::rbtree_best_fit<MutexFamily, VoidMutex, MemAlignment>::size_type = long unsigned int]:
Assertion `priv_end_block() == end_block' failed.

[igaztanaga]

Reviewing old bugs. Heap memory does not support overaligned types because the heap memory returned by operator new is not guaranteed to be suitable for over-aligned types. I will add this to the documentation.