Truncated results when using GROUP BY on a high cardinality column

## Description

`userid` is an INT column with a cardinality of >8M, from 1 to 8,XXX,XXX

TABLE has total of 3,192,209,696 rows.

TABLE has 70 segments distributed over 4 servers running on Pinot with Kubernetes, 10 cores and 35GB memory each (16GB xms/xmx). Using SSD for disk.

## Queries to replicate issue

`SELECT DISTINCTCOUNT(userid) FROM $TABLE WHERE userid BETWEEN 7000001 AND 8000000`

yields 1,000,000 rows as expected

`SELECT userid, count(*) FROM $TABLE WHERE userid BETWEEN 7000001 AND 8000000 GROUP BY userid LIMIT 10000000`

yields 883,574 results

`SELECT userid, count(*) FROM $TABLE WHERE userid BETWEEN 7000001 AND 8000000 GROUP BY userid LIMIT 10000000 OPTION(minServerGroupTrimSize=-1)`

yields 883,574 results

## Configurations

```
# Explicitly set
pinot.server.query.executor.num.groups.limit=10000000 # 10 million

# Not set (using default)
pinot.broker.enable.query.limit.override
pinot.broker.query.response.limit
```

### `index_map`

```
userid.dictionary.startOffset = 0
userid.dictionary.size = 2678308
userid.forward_index.startOffset = 2678308
userid.forward_index.size = 5356608
```

### `metadata.properties`

```
column.userid.cardinality = 669575
column.userid.totalDocs = 48582356
column.userid.dataType = INT
column.userid.bitsPerElement = 20
column.userid.lengthOfEachEntry = 0
column.userid.columnType = DIMENSION
column.userid.isSorted = true
column.userid.hasNullValue = false
column.userid.hasDictionary = true
column.userid.textIndexType = NONE
column.userid.hasInvertedIndex = true
column.userid.hasFSTIndex = false
column.userid.hasJsonIndex = false
column.userid.isSingleValues = true
column.userid.maxNumberOfMultiValues = 0
column.userid.totalNumberOfEntries = 48582356
column.userid.isAutoGenerated = false
column.userid.minValue = 7296465
column.userid.maxValue = 8119811
column.userid.defaultNullValue = -2147483648
```