Add a copy of CircularWindowIterators to the ClientNotificationData of the DataReadyQueue

When inserting a new ID to the queue, create a real copies of all CircularWindowIterators which participate in that 'client update' and save these copies in the ClientNotificationData.

• It would not be required any more to validate windows after each push (CircularBufferManagerBase::validateWindows)
• The client which copies the data from the window can do the validation
• Since we don't know when/if clients will act on new queue elements, the Lifetime of the new CircularWindowIterator copies would be bound to the length of the used DataReadyQueue. When the oldest queue element is dropped, as well is CircularWindowIterators will be released
• Windows would not need to have a valid_ flag anymore

From #31 (closed):

When readout is complete we need to determine if the data is still valid. This would be done by comparing the write head & wraparound counter at the time before sending the window and at the time of readout completion. This way we can detect validity of the data.

Generally agreed. I think after #58 (closed) it should even be sufficient to only validate after readout completion.

This does mean that we need to synchronize the access to the write head (iterator) and the wraparound counter (n_data_pushed_total_) and as such a mutex.

I think we even can prevent that by incrementing n_data_pushed_total_ inside 'CircularBuffer*::push( ... )' operations before doing the actual memcpy. n_data_pushed_total_ would need to be atomic for that, I suppose.

So when reading n_data_pushed_total_ after readout completion and the check still passes, we could be sure that no samples got overwritten during the readout.

As far as I can see, the only drawback would be, that it theoretically can happen that our window will be flagged invalid, even though the memcpy was not done yet. However, that is a rather theoretical use-case, which anyhow would require the write-iterator to first do a full round-trip on the buffer.

You see any problems for doing window validation like that @m.marn ?