Avoid leaking vecs

[nappa85]

This PR avoids using Vecs and then leaking it.

Storage is initialized in a const fashion when creating the allocator, and will live in the same static address, not on the heap. This should increase performances, but I have no benchmarks to demonstrate it (I also understand that performances is not a main goal of this allocator).

I've noticed a logic flaw, and I want to discuss it with you:

• old version was recreating the storage at every start_track, but not resetting allocation_logs_pointer and deallocation_logs_pointer, thus if someone would have called start_track multiple times (obviously respecting the safety guarantees of never calling it concurrently) he could have received the overflow panic even if the buffer wasn't really full
• new version never empties the storage, thus calling start_track multiple times will always keep previous run in memory

A third behavior could be implemented, if desired, emptying the storage and resetting the counters, to allow multiple indepentent runs.

Unrelated, but I would also suggest to use clippy while developing, I've took the initiative to apply some clippy lints on first commit.