AVRO-3869 [chsarp] Fix logical fixed schema support

[paulohumberto42]

What is the purpose of the change

Logical types works only for primitive types, and I can't deserialize an fixed type, like this one:

{
 "type": "fixed",
 "name": "LogicalFixed",
 "logicalType": "decimal",
 "precision": 4,
 "scale": 2,
  "size": 10
}
 

A SchemaParseException is raised ("Undefined name: fixed at 'type'")
This schema is autogenerated by PySpark, from a decimal type, and can't be parsed.
Spark docs: https://spark.apache.org/docs/latest/sql-data-sources-avro.html

Verifying this change

This change added tests and can be verified as follows:

[TestCase("{ \"type\": \"fixed\", \"name\": \"LogicalFixed\", \"logicalType\": \"decimal\", \"precision\": 4, \"scale\": 2, \"size\": 10 }", "decimal", 10)]
public void TestLogicalFixed(string s, string logicalType, int size)
{
    Schema sc = Schema.Parse(s);
    Assert.AreEqual(Schema.Type.Logical, sc.Tag);
    LogicalSchema logsc = sc as LogicalSchema;
    FixedSchema basesc = logsc.BaseSchema as FixedSchema;
    Assert.AreEqual(Schema.Type.Fixed, basesc.Tag);
    Assert.AreEqual(size, basesc.Size);
    Assert.AreEqual(logicalType, logsc.LogicalType.Name);

    testEquality(s, sc);
    testToString(sc);
}

[KalleOlaviNiemitalo]

This pull request changes only parsing, not Schema.ToString(), whose result might still not be interoperable with other libraries. I suppose ToString can be considered out of scope for AVRO-3869 and fixed later in AVRO-3570.