Remove consistency level downgrade retry in CassandraClient

[heesung-sohn]

The current query retry logic applies "CL downgrade" in case of errors.

cdc-apache-cassandra/connector/src/main/java/com/datastax/oss/cdc/CassandraClient.java

Lines 339 to 360 in 0fffe8b

	CompletionStage<Tuple2<AsyncResultSet, ConsistencyLevel>> executeWithDowngradeConsistencyRetry(
	CqlSession cqlSession,
	BoundStatement boundStatement,
	List<ConsistencyLevel> consistencyLevels) {
	final ConsistencyLevel cl = consistencyLevels.remove(0);
	final BoundStatement statement = boundStatement.setConsistencyLevel(cl);
	log.debug("Trying with CL={} statement={}", cl, statement.getPreparedStatement().getQuery());
	final CompletionStage<Tuple2<AsyncResultSet, ConsistencyLevel>> completionStage =
	cqlSession.executeAsync(statement).thenApply(rx -> new Tuple2<>(rx, cl));
	return completionStage
	.handle((r, ex) -> {
	if (ex == null || !(ex instanceof UnavailableException) || consistencyLevels.isEmpty()) {
	log.debug("Executed CL={} statement={}", cl, statement.getPreparedStatement().getQuery());
	return completionStage;
	}
	return completionStage
	.handleAsync((r1, ex1) ->
	executeWithDowngradeConsistencyRetry(cqlSession, statement, consistencyLevels))
	.thenCompose(Function.identity());
	})
	.thenCompose(Function.identity());
	}

the Source sends the query to the Node that sent the "event".
It tries to read at CL.QUORUM, then falls back to CL.ONE

it is really dangerous because it may miss some updates and get an inconsistent view over the row

we should consider to remove this CL downgrade.

[eolivelli]

LGTM