Transform specs are ignored if dimension auto detection is used

The column auto detecting feature does not find the new columns created by transforms.

### Affected Version

All versions of Druid so far (up to 0.15.0)

### Description

Say you have data:
```
{"a":"hello","b":"world"}
{"a":"where","c":"to go"}
```

In a file that lives at: `/Users/vadim/Downloads/test-data.json`

And you ingest it with:
```
{
"dataSchema": {
"dataSource": "Downloads",
"parser": {
"type": "string",
"parseSpec": {
"format": "json",
"timestampSpec": {
"column": "!!!_no_such_column_!!!",
"missingValue": "2010-01-01T00:00:00Z"
},
"dimensionsSpec": {}
}
},
"metricsSpec": [
{
"name": "count",
"type": "count"
}
],
"granularitySpec": {
"type": "uniform",
"segmentGranularity": "DAY",
"queryGranularity": "HOUR",
"rollup": true,
"intervals": null
},
"transformSpec": {
"filter": null,
"transforms": [
{
"type": "expression",
"name": "a_prime",
"expression": "concat(\"a\",'_prime')"
}
]
}
},
"ioConfig": {
"type": "index_parallel",
"firehose": {
"type": "local",
"baseDir": "/Users/vadim/Downloads",
"filter": "test-data.json"
},
"appendToExisting": false
},
"tuningConfig": {
"type": "index_parallel",
"maxRowsPerSegment": null,
"maxRowsInMemory": 1000000,
"maxBytesInMemory": 0,
"maxTotalRows": null,
"numShards": null,
"indexSpec": {
"bitmap": {
"type": "concise"
},
"dimensionCompression": "lz4",
"metricCompression": "lz4",
"longEncoding": "longs"
},
"maxPendingPersists": 0,
"forceGuaranteedRollup": false,
"reportParseExceptions": false,
"pushTimeout": 0,
"segmentWriteOutMediumFactory": null,
"maxNumSubTasks": 1,
"maxRetry": 3,
"taskStatusCheckPeriodMs": 1000,
"chatHandlerTimeout": "PT10S",
"chatHandlerNumRetries": 5,
"logParseExceptions": false,
"maxParseExceptions": 2147483647,
"maxSavedParseExceptions": 0,
"partitionDimensions": [],
"buildV9Directly": true
},
"type": "index_parallel"
}
```

Notice how I am trying to create an `a_prime` column with a transform spec.

The job will work but when you query the data:


You see that there is no `a_prime` column.

I would be great (and would make a ton more sense) if the transforms added themselves to the column list coming from the file.

Start by reproducing the supplied index_parallel ingestion spec with an empty dimensionsSpec and the expression transform that creates a_prime. Trace how column auto detection builds its dimension list relative to the transformSpec. Done means transformed columns such as a_prime appear in the ingested data and are queryable when dimensions are auto detected.

Written by the indexing model from the issue text.