For some reason while doing the "remove blockchain from wallet PR" I
changed the tests around in what I thought was a benign way. But it
meant (apparently) that both parties "test_sync_double_receive" were
using the same "blockchain". This meant that when the blockchain was RPC
they both imported their addresses to it and got each other's results
when syncing. This bugged out the sync and this commit fixes that.
Although somewhat convenient to have, coupling the Wallet with
the blockchain trait causes development friction and complexity.
What if sometimes the wallet is "offline" (no access to the blockchain)
but sometimes its online?
The only thing the Wallet needs the blockchain for is to sync.
But not all applications will even use the sync method and the sync
method doesn't require the full blockchain functionality.
So we instead pass the blockchain in when we want to sync.
- To further reduce the coupling with blockchain I removed the get_height call from `new` and just use the height of the
last sync in the database.
- I split up the blockchain trait a bit into subtraits.
We used to consider a tx input as ours if we had the
tx that creates it in the database.
This commit actually checks if an input is ours before adding
its value to the `sent` field.
Verify the unconfirmed transactions we download against the consensus
rules. This is currently exposed as an extra `verify` feature, since it
depends on a pre-release version of `bitcoinconsensus`.
Closes#352
