Tracking bug for message initialization in C

[clalancette]

In porting efforts we've run into several bugs where the zero initialized behavior is relied upon, causing bugs. Having a zero initialization method as a way to make porting less fragile would be valuable. And likely can be used in other cases when the optimization of using uninitialized constructors are not as important.

This is a tracking bug to deal with field initialization of generated message code in C (originally discussed in ros2/geometry2#25).

[dirk-thomas]

Revert ros2/rcl#277 when this has been implemented.