Adding element to array causes exception with MongoDB Unwrap SMT [DBZ-710]

Migrated from [DBZ-710](https://issues.redhat.com/browse/DBZ-710)

I tried to push another element into a document with an array, while using the unwrap SMT and Avro serialization:

{code}
# Fine
db.customers.insert([
{ _id : 1005, first_name : 'Bob', last_name : 'Hopper', email : 'thebob@example.com', address : { street : 'main street', zip : 12345 },
grades : [ 123, 456, 789]
}
]);

{code}

{code}
# Crashes
db.customers.update(
{ _id: 1005 },
{ $push: { grades: 42 } }
)
{code}

{code}
org.apache.avro.SchemaParseException: Illegal character in: grades.3
connect_1 | at org.apache.avro.Schema.validateName(Schema.java:1151)
connect_1 | at org.apache.avro.Schema.access$200(Schema.java:81)
connect_1 | at org.apache.avro.Schema$Field.(Schema.java:403)
connect_1 | at org.apache.avro.SchemaBuilder$FieldBuilder.completeField(SchemaBuilder.java:2124)
connect_1 | at org.apache.avro.SchemaBuilder$FieldBuilder.completeField(SchemaBuilder.java:2116)
connect_1 | at org.apache.avro.SchemaBuilder$FieldBuilder.access$5300(SchemaBuilder.java:2034)
connect_1 | at org.apache.avro.SchemaBuilder$GenericDefault.withDefault(SchemaBuilder.java:2423)
connect_1 | at io.confluent.connect.avro.AvroData.addAvroRecordField(AvroData.java:898)
connect_1 | at io.confluent.connect.avro.AvroData.fromConnectSchema(AvroData.java:799)
connect_1 | at io.confluent.connect.avro.AvroData.fromConnectSchema(AvroData.java:652)
connect_1 | at io.confluent.connect.avro.AvroData.fromConnectSchema(AvroData.java:647)
connect_1 | at io.confluent.connect.avro.AvroData.fromConnectData(AvroData.java:324)
connect_1 | at io.confluent.connect.avro.AvroConverter.fromConnectData(AvroConverter.java:75)
connect_1 | at org.apache.kafka.connect.runtime.WorkerSourceTask.sendRecords(WorkerSourceTask.java:220)
connect_1 | at org.apache.kafka.connect.runtime.WorkerSourceTask.execute(WorkerSourceTask.java:187)
connect_1 | at org.apache.kafka.connect.runtime.WorkerTask.doRun(WorkerTask.java:170)
connect_1 | at org.apache.kafka.connect.runtime.WorkerTask.run(WorkerTask.java:214)
connect_1 | at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
connect_1 | at java.util.concurrent.FutureTask.run(FutureTask.java:266)
connect_1 | at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
connect_1 | at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
connect_1 | at java.lang.Thread.run(Thread.java:748)
{code}

The report names no repository file or test path. Start with the MongoDB $push reproduction and the AvroData.fromConnectData/SchemaBuilder stack trace, then trace how the unwrapped array update is converted before Avro serialization. Done means appending an array element no longer throws SchemaParseException in the stated setup.

Written by the indexing model from the issue text.