Cardano Updates

Avoid tmp allocations in acknowledgeTxIds by folding acknowledged txids once
for mempool queueing and refcount updates.

Avoid tmp allocations in receivedTxIdsImpl by updating available/buffered
tx maps in one pass.

Avoid building temporary Maps and Sets in pickTxsToDownload and
filterActivePeers.

If state change wait at most 25ms for additional changes in order to
coalece TX events together.

Cut down on the number of wakeups by only depending on the debouncer.

Merge countRejectedTxs and updateBufferedTx calls into the same atomic
operation so that submitTxsToMempool only bumps the generation number
once.

Move tx counter tracing to drainRejectionThread where it will be run at
most once per second.

Send lists of TXs to the mempool when possible. This mimics the
behaviour of the V1 tx submission.

Use a generation counter to track changes to SharedStateVar and only
call the expensive makeDecisions if a timer has expired or the shared
state has changed.

spanl will silently drop trailing tx's if there is a gap
in the txs to acknowledge sequence. This may happen if acknowledging
tx's which some other peer has downloaded.

Instead of always waiting 10ms and then loop around we check when we can
issue the next request and delay until then.

This fixes a bug where the same tx is enqueued multiple times for
submission into the mempool by the same peer.

cardano topology

ouroboros-network-1.1.0.0 release

Currently the channel-based ones are failing due to the way we terminate the test client/server implem

Avoid tmp allocations in receivedTxIdsImpl by updating available/buffered
tx maps in one pass.

Avoid tmp allocations in acknowledgeTxIds by folding acknowledged txids once
for mempool queueing and refcount updates.

Avoid building temporary Maps and Sets in pickTxsToDownload and
filterActivePeers.

Merge countRejectedTxs and updateBufferedTx calls into the same atomic
operation so that submitTxsToMempool only bumps the generation number
once.