Need timeouts and proper task-cancellation when dealing with trinity's proxied-services

Currently our chain syncer implementation do things like:

await wait_with_token(chaindb.coro_import_block(block), token=self.cancel_token)

Where chaindb is a multiprocessing proxy that runs the actual block importing in a separate process. Since we’re dealing with a separate process that may die halfway through importing the block, or hang completely and never respond, I believe we should always specify timeouts when dealing with them – in the same way we do for network operations.

Also, by wrapping those with wait_with_token() only cancels the outer task and not the one that is actually doing the work, which is created inside coro_import_block() by the await loop.run_in_executor(…) call. Maybe we just need to pass the CancelToken to coro_import_block() and have it use that with wait_with_token() instead of us.

[gsalgado]

#860 addresses part of this

[gsalgado]

@carver actually, they’re not terminated in order: https://github.com/ethereum/py-evm/issues/827#issuecomment-394317506