Snapshot sync

[arkpar]

• Enable snapshot service over IPC
• Sync to the snapshot from other peers

Still left to do:

• Background block downloading
• Snapshot evaluation - don't sync from the snapshot if not too far behind
• Protocol documentation
• Fork block cache (post-snapshot sync after hard fork has no hard-fork block #2043)

[coveralls]

Coverage decreased (-0.06%) to 84.385% when pulling a98733d on pv64-sync into 59f18ab on master.

[rphmeier]

RwLock<Arc<T>> doesn't guarantee or provide anything as Arc only has a Deref implementation

[arkpar]

It is a lock for a shared pointer. It is required to synchronously replace the pointer with another once.

[tomusdrw]

But you can always just clone the Arc instead of accessing the shared pointer. Maybe it's worthing wrapping Arc<BlockChain> into a structure with Deref but without Clone to prevent such situations?

[rphmeier]

I would favor giving Client full ownership of BlockChain, only loaning it out temporarily to the TraceDB.

[arkpar]

@tomusdrw I need to replace the pointer so that it points to another object. This is not the same as cloning Arc

[arkpar]

@rphmeier I'll leave this to another PR. This one is too big already

[tomusdrw]

Yeah, I understand, but the point is that even if you replace the pointer someone might be using the old one anyway (because after getting a lock you can just clone the Arc and release the lock).

[arkpar]

Ah, ok. I'm sure that's not happening cause this pointer is not exposed out of the client. It is only used in TraceDB and we should really just pass a reference there as @rphmeier suggested.

[coveralls]

Coverage decreased (-0.08%) to 84.364% when pulling 2b3df3b on pv64-sync into 59f18ab on master.

[coveralls]

Coverage decreased (-0.1%) to 84.572% when pulling da2940b on pv64-sync into 4e466f0 on master.