points not deduplicated from overlapped shards

__Steps to reproduce:__
1. start influxdb server

```
$ docker run --rm --detach --name influxdb -e INFLUXDB_MONITOR_STORE_ENABLED=false -e INFLUXDB_DB=foo influxdb:1.8.10
```
2. run following statements
```
docker exec -it influxdb influx -database foo -precision s
Connected to http://localhost:8086 version 1.8.10
InfluxDB shell version: 1.8.10
> insert cpu,host=abc value=111 1641254400
> alter retention policy autogen on foo shard duration 10d
> insert cpu,host=abc value=222 1640995200
> insert cpu,host=abc value=333 1641254400
> select * from cpu
name: cpu
time host value
---- ---- -----
1640995200 abc 222
1641254400 abc 111
1641254400 abc 333
> show shards
name: foo
id database retention_policy shard_group start_time end_time expiry_time owners
-- -------- ---------------- ----------- ---------- -------- ----------- ------
2 foo autogen 2 2021-12-27T00:00:00Z 2022-01-06T00:00:00Z 2022-01-06T00:00:00Z
1 foo autogen 1 2022-01-03T00:00:00Z 2022-01-10T00:00:00Z 2022-01-10T00:00:00Z
>
```

Two points with the same timestamp are returned, which is not expeccted

The issue names no source file or test. First reproduce the shard-overlap scenario with the provided Docker and InfluxQL commands, then trace the query results across overlapping shards; done means points sharing a timestamp are deduplicated as expected.

Written by the indexing model from the issue text.