Make FinalityFlow do more, and be used more consistently.
Add a new waitForLedgerCommit API that is intended to be used at the end of flows, or at any other point where a flow wants to wait for a transaction to finalise (but the finalisation flow is being done by someone else).
Update the docs a bit.
* Move merkle building extension functions on wire tx to WireTransaction class.
* Add timestamp, notary, transaction type and signers to wire transaction id calculation.
* Change construction of MerkleTree from duplicating last node on a given level to padding leaves' list with zero hash to size of the nearest power of 2 - so we always have a full binary tree.
The problem was that it was possible to construct 2 different transactions with the same ids. Trick worked for txs having number of leaves that were not power of 2.
* Update tear-offs documentation and diagrams to reflect changes in construction of Merkle trees - padding with zero hashes and including all WireTransaction fields in id computation.
* Change in filtering API of WireTransaction for partial Merkle trees calculation.
Instead of many filtering functions over a transaction only one needs to be provided.
Additional change to check and verification of FilteredTransaction.
* IRS demo change. Make filtering function a protected method of RatesFixFlow class.
Comment on situation when capturing too much scope and connected problems with checkpointing.
Change oracle and tear-offs documentation.
The IRS demo uses names of parties to identify them in flows, which introduces problems when a node may not
know who a party is, as well as being a poor choice of key generally. This switches to use the owning key
instead.
Change Party instances to be uniquely identified by the owning key, without taking into account name.
This requires that mock node key generation is reworked so that keys for services and the node itself
are distinct, otherwise the network map service cannot differentiate them.
Signed-off-by: Ross Nicoll <ross.nicoll@r3.com>
This deprecates the existing composition clauses and adds new better named versions,
as well as changing 'AnyOf' to require at least one matching subclase (better matching
the name of the clause).
