Peter M. Groen / FlashMQ

When there are many sessions, the copying action can actually take some
time, and memory. A vector is faster and uses (a little) less memory.

This is a theoretical fix, without benchmarks to support it.

The only mutable session data of a client is QoS related, so when we're
copying sessions (for saving them), we need to lock the QoS data,
because that gets modified from active client traffic in worker threads.

Note: not super well tested at this point, nor was I ever able to
trigger actual errors despite long stress testing, so it's a theoretical
fix.

Empty leaves weren't removed, causing increase in memory use.

This method incurs no extra CPU load when messages aren't dropped.

One was confirmed: writing an mqtt packet into a client that
disconnected after checking the weak pointer for validity.

The rest made sense to change as well.

I have no explanation for it, but it happens (with QMQTT, should it
matter).

It was a mistake that it ran inside the timer thread, which came to
light by asserting the thread local authentication pointer (see a few
commits ago).

There were bugs in which authentication object was used when, causing
threadings bugs. Instead of getting from the 'sender', we can just store
a thread local pointer.

Files are simple serialized bytes prefaced by lengths. File is hashed to
verify integrity. This was also a good way preventing unexpected errors
when trying to crash the parser by having it load a different file.

This change includes some refactoring that was necessary:

- It 'fixes' looking at the wrong thread's authentiction. This is still
  wrong though. It will be fixed by a thread local pointer in the next
  commit.
- Deadlocks with yourself are handled in rwlockguard.
- QoSPacketQueue is now a class.
- Probably other tweaks.

It didn't clean up the old ones, causing unpredictable problems.

This fixes side effects when system time changes.

This performs a whole lot better.

Three parts to it:

- Start the app fresh per test. This avoids annoyances like getting
  retained messages on subscribe, messing up the test.
- Waiting for suback was apparently necessary.
- Because the Qt event loop was given time, waiting for publishes was
  sometimes pointless because it had already arrived. So, checking the
  receive list first.

Fix version bug.

Also include a few stats.