Message dumping is slow for big files

Currently all the dump files are valid JSON files, which means that for every message to be dumped, this happens:

• Read and parse the full dump-file
• Append the new dump
• Re-write the entire file

Obviously, this becomes incredibly slow for bigger files.

One possibility for fixing this is to work line-based. So, it's not a valid JSON file anymore (at least with more than one line), but each line is on itself. Consequently, only one line needs to be appended. That's faster and less error-prone.

[e3rd]

• We shouldn't dump the same line twice. (exceeding dumping #994)