Does insert operation also create index after write?

Hi,

I am using Hudi 0.15.0, I have following sql:

```
set hoodie.spark.sql.insert.into.operation=insert;
set hoodie.datasource.write.insert.drop.duplicates=false;
set hoodie.datasource.write.insert.dup.policy=none;
set hoodie.combine.before.insert=false;

CREATE TABLE IF NOT EXISTS hudi_cow_20260102_06 (
a INT,
b INT,
c INT
)

USING hudi

tblproperties(
type='cow',
primaryKey='a',
hoodie.datasource.write.precombine.field='c',
hoodie.index.type='BLOOM',
hoodie.index.bloom.num_entries='20',
hoodie.bloom.index.filter.dynamic.max.entries='25'
);

insert into hudi_cow_20260102_06(a,b,c) values(1,2,3),(1,4,7),(1,3,6);
```

There are 3 records in hudi_cow_20260102_06 after the insertion (all the three record's record key is 1)

When I look at the parquet footer, I see the bloom filter is created in in the parquet footer, which means, bloom index is created( `hoodie_bloom_filter_type_code: DYNAMIC_V0` in the footer).

I have thought that no index will be created after duplicates allowed insertion, because there maybe duplicates for `insert`, if two records have the same record key, which record would the index refer to?

I am using the following command

```

[hadoop@hadoop ~]$ hadoop jar software/parquet-cli-1.14.1-runtime.jar meta hdfs:///user/hive/warehouse/hudi_cow_20260102_06/e6ffba4b-0b30-4333-a851-7bef6dcc9cb0-0_0-612-617_20260102160515023.parquet

File path: hdfs:///user/hive/warehouse/hudi_cow_20260102_06/e6ffba4b-0b30-4333-a851-7bef6dcc9cb0-0_0-612-617_20260102160515023.parquet
Created by: parquet-mr version 1.12.2 (build 77e30c8093386ec52c3cfa6c34b7ef3321322c94)
Properties:
hoodie_bloom_filter_type_code: DYNAMIC_V0
org.apache.hudi.bloomfilter: /////wAAAB4BAAADXwAAABQAAAADAAAAAf////8AAAAeAQAAA18AAAAAEAAAAEAAAAQAEIAAAAAAAQAAAAIAAAAAIAAAAAAIAAAIAAAAAAAAAAAAyAAAgAIAAAAAAAAAAIAAABBABAAAAAAAAAQAIAAQAAAAAABAABAAAAAAAAAAAIAAAAAAAAAAAAEAAggAEAA=
hoodie_min_record_key: 1
parquet.avro.schema: {"type":"record","name":"hudi_cow_20260102_06_record","namespace":"hoodie.hudi_cow_20260102_06","fields":[{"name":"_hoodie_commit_time","type":["null","string"],"doc":"","default":null},{"name":"_hoodie_commit_seqno","type":["null","string"],"doc":"","default":null},{"name":"_hoodie_record_key","type":["null","string"],"doc":"","default":null},{"name":"_hoodie_partition_path","type":["null","string"],"doc":"","default":null},{"name":"_hoodie_file_name","type":["null","string"],"doc":"","default":null},{"name":"a","type":["null","int"],"default":null},{"name":"b","type":["null","int"],"default":null},{"name":"c","type":["null","int"],"default":null}]}
writer.model.name: avro
hoodie_max_record_key: 1
Schema:
message hoodie.hudi_cow_20260102_06.hudi_cow_20260102_06_record {
optional binary _hoodie_commit_time (STRING);
optional binary _hoodie_commit_seqno (STRING);
optional binary _hoodie_record_key (STRING);
optional binary _hoodie_partition_path (STRING);
optional binary _hoodie_file_name (STRING);
optional int32 a;
optional int32 b;
optional int32 c;
}

Row group 0: count: 3 248.33 B records start: 4 total(compressed): 745 B total(uncompressed):551 B
--------------------------------------------------------------------------------
type encodings count avg size nulls min / max
_hoodie_commit_time BINARY G _ R 3 37.00 B 0 "20260102160515023" / "20260102160515023"
_hoodie_commit_seqno BINARY G _ 3 26.67 B 0 "20260102160515023_0_0" / "20260102160515023_0_2"
_hoodie_record_key BINARY G _ R 3 31.67 B 0 "1" / "1"
_hoodie_partition_path BINARY G _ R 3 31.33 B 0 "" / ""
_hoodie_file_name BINARY G _ R 3 53.00 B 0 "e6ffba4b-0b30-4333-a851-7..." / "e6ffba4b-0b30-4333-a851-7..."
a INT32 G _ R 3 31.33 B 0 "1" / "1"
b INT32 G _ 3 18.67 B 0 "2" / "4"
c INT32 G _ 3 18.67 B 0 "3" / "7"
```

```

Reproduce the Hudi 0.15.0 SQL using the stated insert settings, then inspect the generated Parquet footer with parquet-cli as shown. Compare the observed Bloom index metadata with Hudi's documented insert and index behavior; done means clearly explaining whether this is expected and identifying any documentation gap.

Written by the indexing model from the issue text.