shorten dmesg.log

[Edgar Fuß]

> Would you mind taking a look at the logtail patch that's been sitting in
> experimental for almost forever already?
I don't see how this can possibly work on something like the kernel message buffer. Its contents will rotate once it runs full so the offset noted by logtail will no longer be correct.
I think the only way of handling that is putting a marker into the buffer that will also rotate. (In case it rotates out, we take everything we can get, which is just what would be done now.)