LibCore: Avoid a big malloc in CIODevice's internal buffering.

This heap allocation was totally avoidable and can be replaced by a stack
buffer instead. Dodges a bunch of mmap() traffic.

LibCore/CIODevice.cpp

@@ -146,8 +146,8 @@ bool CIODevice::populate_read_buffer()
 {
  if (m_fd < 0)
  return false;
- auto buffer = ByteBuffer::create_uninitialized(PAGE_SIZE);
- int nread = ::read(m_fd, buffer.pointer(), buffer.size());
+ byte buffer[1024];
+ int nread = ::read(m_fd, buffer, sizeof(buffer));
  if (nread < 0) {
  set_error(errno);
  return false;
@@ -156,8 +156,7 @@ bool CIODevice::populate_read_buffer()
  set_eof(true);
  return false;
  }
- buffer.trim(nread);
- m_buffered_data.append(buffer.pointer(), buffer.size());
+ m_buffered_data.append(buffer, nread);
  return true;
 }