Use unbuffered channels for lambda and agent data

[lahsivjar]

Currently, we are using a buffered channel by default for lambda and agent data (ref). This is not required anymore as we already have a buffer in accumulator. The channel based buffer may cause additional issues with lifecycle management of the extension as it is possible that the buffer has a lot of data but the function doesn't have time to process that data and has been frozen due to inactivity.