Add a deduplication utility for raw buffers

[jkflying]

This is a sample implementation of a deduplication tool. Not yet integrated into mavlink-router message handling or configuration.

[jbeyerstedt]

With the change above (and some merge-conflict resolving to make it work with upstream) it seems to work just fine, even with our special multi-modem/ redundant-link situation.

(But the auterion repo and upstream are already quite different. I think, that we need to get your changes upstream as soon as possible and not continue with two diverging code-bases.)

[jbeyerstedt]

I'm currently trying to add a config parameter for the dedup period, but mavlink-router's architecture is not really cooperating. It's not really in a good place now, but I didn't want to add another method to the Mainloop just for configuring the de-duplication. See the latest commit of https://github.com/jbeyerstedt/mavlink-router/commits/feature/de-duplication.

P.S. Did I say, that I want to re-do the whole config stuff?
P.P.S. Sorry for ranting in your PR.

[jbeyerstedt]

We have some issues with missing data after letting this run for some time. Especially the heartbeat will not get through mavlink-router any more. I guess, the issue has to do with the code I commented on.

And by the way, the output of the duration_cast is a signed integer, but even "just" using half of the 32 bit value range will last for 24 day until an overflow. (Should mavlink-router just quit in case of an integer overflow and be re-started by systemd?)
But the more interesting issue is, that the timestamp value starts at about 980 million milliseconds, which seems a bit odd. It's not the right value for a signed to unsigned conversion. Which lead me to the question, where start_time is initialized!?

[jbeyerstedt]

This might get into upstream master before the next tagged release of mavlink-router with PR mavlink-router#323. We have done some major refactoring and cleanup in the master branch lately, so I had to change some more things to get it rebased.