mixed index with SolrCloud returns inconsistent results

Nobody has claimed this yet.

Using SolrCloud with 2 nodes, and a mixed index with 2 shards and replication factor of 2, the streaming result can end up with duplicates.

This behavior was not witnessed in these configurations:
* Solr Standalone
* SolrCloud with 1 node, mixed index with 1 shard, RF=1
* SolrCloud with 1 node, mixed index with 1 shard, RF=2
* SolrCloud with 1 node, mixed index with 2 shard, RF=1
* SolrCloud with 1 node, mixed index with 2 shard, RF=2
* SolrCloud with 2 nodes, mixed index with 1 shard, RF=1
* SolrCloud with 2 nodes, mixed index with 1 shard, RF=2
* SolrCloud with 2 nodes, mixed index with 2 shard, RF=1

- Version: JanusGraph 0.3.1, 0.3.3, 0.4.0
- Storage Backend: berkeleyje
- Mixed Index Backend: solr 7.4.0, 7.7.2
- Steps to Reproduce:

janusgraph.properties

```
storage.backend=berkeleyje
storage.directory=db/berkeleyje

index.search.backend=solr
index.search.solr.mode=cloud
index.search.solr.zookeeper-url=localhost:2181
index.search.solr.configset=janusgraph_configs
index.search.solr.num-shards=2
index.search.solr.replication-factor=2
index.search.solr.max-shards-per-node=4
# default max-result-set-size is 50
index.search.max-result-set-size=2
```

Schema and index definition

```
final PropertyKey name = mgmt.makePropertyKey("name").dataType(String.class).cardinality(Cardinality.SINGLE).make();
final JanusGraphIndex nameIndex = mgmt.buildIndex("nameIndex", Vertex.class).addKey(name).buildMixedIndex(indexName);
```

Graph initialization (10 vertices)

```
for (int i=0; i<10; i++) {
g.addV().property("name", "solr").next();
}
g.tx().commit();
```

Query against mixed index

```
g.tx().rollback();
List list = g.V().has("name", "solr").toList();
g.tx().rollback();
Set set = new TreeSet<>();
for (Vertex v : list) {
set.add(v);
}
System.out.println(list.size() == set.size());
```

The query will return sometimes return fewer than 10 vertices. If you increase the value of `index.search.max-result-set-size` to something larger than the expected number of results, then the mixed index query returns the correct number.

I observed that the query sent to Solr has no ordering involved. I suspect that because there is no sort order, the results are returned in an indeterminate order. When the results are coming back from multiple nodes across multiple shards, the likelihood of getting a duplicate increases.

Start by reproducing the issue with the supplied janusgraph.properties configuration, mixed-index definition, and query against a two-node, two-shard, RF=2 SolrCloud setup. Compare the returned vertices with the expected 10 and inspect the generated Solr query, including its lack of ordering; done means the query consistently returns all vertices without duplicates in the reported configuration.

Written by the indexing model from the issue text.