Better graphql performance

[quentinovega]

##Context

The current GraphQL implementation has Fetcher/DeferredResolver for batching, but it is underutilized:

• Only 4-5 fields use .defer() across the 3 existing fetchers (teamsFetcher, apisFetcher, usersFetcher)
• Approximately 15 direct findById calls in the resolvers bypass batching (N+1)
• The existing fetchers perform Future.sequence(ids.map(findById)) — each ID generates an individual DB query
• No tenantsFetcher even though there are 6 non-batched tenant searches

On a large cluster of objects, this can consume all database connections and block other users.

Actions

1. Implement batch queries in the DataStore

Expose the findByIds(ids: Seq[Id]) methods using the Seq[Ids] provided to replace the N individual queries with a single one.

2. Update existing Fetchers

Replace Future.sequence(ids.map(findById)) with a call to findByIds in teamsFetcher, apisFetcher, and usersFetcher.

3. Create new fetchers

Add a dedicated fetcher for the 6 tenant searches that are currently not batched. Think to create another fetchers (for plans or another items)

4. Convert direct resolvers to .defer()

Migrate the approximately 15 findById calls in the resolvers to .defer() calls on the appropriate fetcher.

5. (Optional) Configure maxBatchSizeOpt on the fetchers

Limit batch sizes to avoid excessively large $in queries.

Affected Files

• app/fr/maif/daikoku/domain/SchemaDefinition.scala — fetchers and resolvers
• DataStore / repos layer — adding findByIds methods