`job_logs_buffered` streaming API behavior improvement

[pankgeorg]

Some unintuitive behaviours of the API are:

• the callback doesn't run on existing logs. that means that I need to call my functions twice
• there are no markers that delineate the changes; i.e. what logs are unprocessed

A backwards compatible solution could be to wrap the function with a helper like:

job_logs_appy(callback)

and make sure the callback gets triggered serially (and in line) to all existing and future logs.