Clarify that NULL blocks forever in rmw_wait

[sloretz]

The documentation says rmw_wait() blocks forever if the timeout is negative, but the timeout cannot be negative because rmw_time_t uses unsigned types.

typedef struct RMW_PUBLIC_TYPE rmw_time_t
{
  uint64_t sec;
  uint64_t nsec;
} rmw_time_t;

If ros2/rcutils#79 leads to rclutils_time_point_value_t using a signed type then it may make sense to propagate signed-ness down to rmw, but either way this documentation should still say it blocks when the time pointer passed in is NULL.

CI Linux, only testing rmw for any linter tests since this is a documentation change only

• Linux

[wjwwood]

This is the simplest change, but long term we should probably switch to rcutils_duration_value_t and remove the rmw_time_t.