In-LDM event squashing (instead of an output buffer)

[KB1RD]

In my various implementations using the LDM, I often hold events that cannot be sent immediately (in the case of Matrix, this is due to sending limits) in an output buffer. When consecutive operations are sitting in the buffer waiting to be sent, they are merged together if possible to reduce the number of events sent.

However, this requires an algorithm to merge nodes together and as I started improving it, it began to look more and more like the LDM itself. In that case, why not just grab the nodes right out of the LDM itself? The insertion and removal methods could be modified to additionally return an object specifying the relevant region of nodes where the insertions/removals originated from. When the events being sent end up stuck in a queue, these regions only are merged together if possible. Then, the LDM is asked for the insertions/removals in that region. This effectively offloads squashing to the LDM and potentially opens up an opportunity to share the current state of the document with other peers.

[KB1RD]

This would require at least a partial implementation of #4