This PR aims to optimize Knowledge by appropriately setting SQL indexes on the fields that are frequently used to join tables and filter out records. The new indexes should help reducing the server response times.

For the `knowledge.article` model:

1. Remove the `is_article_item` index

Setting an index on a boolean field is generally speaking not recommended as the field has only two possible values and does not allow us to discriminate the records efficiently. Indeed, the index will be used only when searching for an uncommon boolean value. In the other cases, the index does not provide much benefit compared to a full table scan. PostgreSQL will then favor another index that will better reduce the search space of the query.

2. Remove the `category` index

The index set on the `category` column is almost never used as we never search for articles based on their categories and we never join tables based on that value. The index could be used when the user applies a search filter in the list view but, in that case, PostgreSQL could use another index that better discriminate the records.

3. Add index on `parent_id`

The `parent_id` field is heavily used in the user's permission computation. Adding an index for that field will ensure that we can quickly explore the article hierarchy and compute the user's permissions efficiently.

For the `knowledge.article.member` model:

1. Add index on `article_id`

The `article_id` field is heavily used in the user's permission computation. Adding an index for that field will ensure that we can quickly compute the user membership on the article.

task-4046060