count distinct and from_unixtime return different result compared to Vanilla Spark

### Backend

VL (Velox)

### Bug description

When using count distinct and from_unixtime, the result shows difference with Vanilla Spark.

Below is the reproduce case:
select a1
,count(1)
, 1000/count(distinct if(a1=a2, b, null)) as send_check_avg_cnt_sd
, from_unixtime(unix_timestamp(max(d),'yyyyMMdd'),'yyyy-MM-dd') as last_wb_detail_view_day,max(d)
, count(distinct if(a1=a2 and c in ('4'), b, null)) as consignor_check_waybill
from count2
group by a1 order by a1 limit 100 ;

### Spark version

Spark-3.2.x

### Spark configurations

_No response_

### System information

_No response_

### Relevant logs

_No response_