feat: add aggregate query support (GROUP BY + aggregate functions)

[Copilot]

Summary

Adds Q.aggregate(schema, aggregateMap) — a new query operator that performs GROUP BY + aggregate functions at the database level (Memory, SQL, Cosmos) instead of fetching all rows and grouping in memory.

Motivation

Replaces patterns like:

const orders = yield* orderRepo.query(Q.where(...), Q.project(...))
const byShipment = new Map()
for (const order of orders) {
  // manual group-by + counting
}

With a single DB-level aggregate query:

const result = yield* orderRepo.query(
  Q.where("shippingDate", "gte", today.toISOString()),
  Q.where("state._tag", "neq", "cancelled"),
  Q.aggregate(
    S.Struct({ cdcCity: CDCCity, shippingDate: S.DateFromString, picks: S.NonNegativeInt, ... }),
    {
      cdcCity: Q.agg.field("cdcAddress.city"),
      shippingDate: Q.agg.field("shippingDate"),
      picks: Q.agg.countWhen((q) => q.pipe(Q.where("state._tag", "in", ["initial", "valid", "picking"]))),
      packs: Q.agg.countWhen((q) => q.pipe(Q.where("state._tag", "eq", "picked"))),
      packed: Q.agg.countWhen((q) => q.pipe(Q.where("state._tag", "eq", "packed"))),
      done: Q.agg.countWhen((q) => q.pipe(Q.where("state._tag", "eq", "closed")))
    }
  )
)

New API

agg DSL namespace:

• agg.field(path) — group-by field reference (doc path → output alias)
• agg.count() — COUNT(*)
• agg.countWhen(operation) — COUNT of rows matching filter
• agg.sum(field) — SUM of a numeric field
• agg.min(field) / agg.max(field) — MIN/MAX of a field

aggregate(schema, aggregateMap) — query operator using the above

Changes

File	Change
src/Model/query/new-kid-interpreter.ts	AggregateIrExpression type, AggregateIrItem, extend Result, handle "aggregate" mode
src/Model/query/dsl.ts	AggregateExpression, AggregateMap, agg const, aggregate() function
src/Store/service.ts	Extend FilterArgs.select with { key, path } and { key, aggregate } item types
src/Store/Memory.ts	In-memory GROUP BY + aggregation in memFilter
src/Store/SQL/query.ts	GROUP BY + COUNT(CASE WHEN ...) / SUM / MIN / MAX
src/Store/Cosmos/query.ts	GROUP BY + SUM(IIF(...)) for conditional counts
src/Model/Repository/internal/internal.ts	"aggregate" mode dispatch — bypass id/_etag augmentation, decode raw rows
test/sql-store.test.ts	4 new aggregate integration tests
test/cosmos-query.test.ts	3 new aggregate query builder tests
test/query.test.ts	3 new memFilter aggregate unit tests

[patroza]

@copilot Is there some overlap now between especially agregrate and projectComputed?

[Copilot]

@copilot Is there some overlap now between especially agregrate and projectComputed?

There is some conceptual overlap (both compute values in the query layer), but they solve different shapes: projectComputed is per-row enrichment (1 output row per entity), while aggregate is grouped/collapsed output (GROUP BY, 1 row per group). So projectComputed is for entity projections; aggregate is for reporting/rollups.