* get rid of `float_components` class
  * fix GCC 8 `-Wclass-memaccess` warnings as a side effect
* assemble, disassemble float types directly with the aid of `memcpy`,
  bitwise integer operations