MyPupils: Replace AISearch pupil hydrate with a CosmosDb/PostGres impl

[aahmed-dfe]

Currently AggregatePupilsApplicationService uses AISearch to hydrate Pupil information, once we have the List from CosmosDb.MyPupils.

We should replace this hydration from AISearch to a CosmosDb impl to

1. remove depedency on AISearch which could reduce costs if AISearch is removed
2. Hydrate entirely through CosmosDb than multiple infrastructure parts.

This will also allow us to Server-side sort/paginate natively via SQL (this can be incremental applied, as currently we Client-side sort on a limited list).

[aahmed-dfe]

Issues with backing onto CosmosDb.

1. 2 Containers being sorted into a single List<Pupil>

There are 2 containers which both have the same PupilMatchingRef partitionKey. I assume this is ingested from upstream.

Given they are separate, the most likely way is to merge_sort between both, as I iterate. Both repositories would expose a Iterator, that would be mapped into the ApplicationModel as the Iterator progressed the underlying DTO.

Both repositories would receive the QueryModel to implement into a SortBy.

2. Offset / Skip is not built into CosmosDb.

The pagination model uses a ContinuationToken and allows paging like IAsyncPageable by passing this continuation token. So I'd still be client-side paging through results.

---

Given a move to Postgres is on the horizon, look at building out a single pupil-read model that could be used across Search and MyPupils.

Clients always send us a UPN for NPD/FE Search and MyPupils.

Likely;

Primary table; PK PupilMatchingRef, Pupil information
Link table PupilMatchingRef : UPN
{N linked tables off FK PupilMatchingRef with relevant data}

View/projection that we consume pupil-read, pupil-download

Dapr to Query.
Resolve through link table

Will need to bring in postgres extensions for FacetedFilter, FuzzyMatch