[Enhancement]: Remove the deduplication logic

[xiaocai2333]

Is there an existing issue for this?

• I have searched the existing issues

What would you like to be added?

Remove the deduplication logic from the read/write patch, and allow duplicate primary keys in query results.

In the future, we will introduce a deduplicated collection that handles duplicates during writes.

Why is this needed?

Milvus's deduplication is currently incomplete. For example, count(*) does not deduplicate. For consistency, query results also retain duplicates.

Anything else?

No response

[xiaofan-luan]

This could be very important for debugging.

We will rollout the primary key deduplication in 3.0