Performance: Stream query results instead of materializing full lists

[mikemcdougall]

Problem

Query endpoints materialize full result sets in memory instead of streaming, increasing memory pressure and latency for large datasets.

Evidence (current behavior)

• OGC Features: featureStore.QueryAsync then result.Items → .ToArray() in src/Honua.Server/Features/OgcFeatures/FeaturesEndpoints.cs:309 and src/Honua.Server/Features/OgcFeatures/FeaturesEndpoints.cs:310.
• FeatureServer: QueryAsync in src/Honua.Server/Features/FeatureServer/FeatureServerHandler.cs:1434 and full formatting on result.Items in src/Honua.Server/Features/FeatureServer/FeatureServerHandler.cs:219 / src/Honua.Server/Features/FeatureServer/FeatureServerHandler.cs:227.
• OData: QueryAsync then queryResult.Items.Select(...).ToArray() in src/Honua.Server/Features/OData/ODataEndpoints.cs:398 and src/Honua.Server/Features/OData/ODataEndpoints.cs:409.
• Storage layer builds lists: ExecuteSelectQueryAsync accumulates List<Feature> in src/Honua.Postgres/Features/FeatureStore/Services/FeatureDataAccess.cs:83.

Direction

• Use existing IStreamingFeatureStore (src/Honua.Core/Features/FeatureStore/Abstractions/IStreamingFeatureStore.cs:11).
• Stream features via IAsyncEnumerable + DbDataReader and serialize incrementally (e.g., Utf8JsonWriter writing arrays).
• Keep count as a separate query or optional when returnCountOnly is false.

Tasks

• Add streaming response writers for FeatureServer / OGC / OData query endpoints.
• Wire endpoints to IStreamingFeatureStore when available.
• Ensure JSON/GML serialization supports incremental writing.
• Add tests covering large results and returnIdsOnly/count paths.

Acceptance criteria

• No full in‑memory materialization for large queries in core endpoints.
• Streaming path uses existing IStreamingFeatureStore and Postgres streaming methods.