vec: remove capacity == 0 early return

[wmww]

*_with_capacity(0) allocates, and malloc may return a valid pointer that needs to be freed, even when 0 size is specified. This behavior might be valgrind-specifiec? but malloc(0) is indeed implemenation-defined. An alternative would be to have cap ? _i_new_n(_m_value, cap) : NULL in the *_with_capacity implementation, and a few other places.

[tylov]

Thanks for making me aware. I will accept the commit, but will modify the code to test if the data pointer is NULL instead of removing the test. This is to avoid calling the user-provided free-function with NULL argument. free() does allows this, but the user may assume there is a non-NULL pointer passed, which may lead to UB (like memset(NULL, 0, 0)). Next commit updates the other relevant containers the same way.