Fix leak in the buffer vectorMembers when reading SIO frames

[jmcarcell]

BEGINRELEASENOTES

• Fix leak in the buffer vectorMembers when reading SIO frames

ENDRELEASENOTES

The leak happens because at the time of this emplace_back, there is already an element in the vector with the string corresponding to that vector member. Then, the indexing will only find the first instance and this second one is leaked, so the fix is to check before and only if it isn't there add it. I think this is added for the first time in https://github.com/AIDASoft/podio/blob/master/python/templates/SIOBlock.cc.jinja2#L92

This helps with the read_frame_sio test. Before:

SUMMARY: AddressSanitizer: 12714 byte(s) leaked in 482 allocation(s).

After:

SUMMARY: AddressSanitizer: 11786 byte(s) leaked in 445 allocation(s).

I'm not sure if we need a test since read_frame_sio would be the one, once it doesn't have any leaks. Otherwise to reproduce we just have to read a frame with a collection with a vector member.

I did this one because after this I have a version of #583 without leaks.

[tmadlener]

IIUC all of this can go, because the buffers at this point already have these vector members set correctly from here:

podio/python/templates/macros/collections.jinja2

Lines 204 to 206 in 17a1201

	{% for member in VectorMembers %}
	readBuffers.vectorMembers->emplace_back("{{ member.full_type }}", new std::vector<{{ member.full_type }}>);
	{% endfor %}

No idea, why this got to here and why it seems to pass all sanitizers and tests. (probably because it's not properly tested).

[jmcarcell]

Yep, that looks to be correct, no more leaks related to that.