Support cross-partition aggregate queries

### Feature Summary

Add production cross-partition COUNT, SUM, MIN, MAX, and AVG query execution to the Cosmos DB Rust driver.

### Feature Description

Merge Gateway rewritten aggregate payloads across partitions and produce the expected public query shape.

Validation and acceptance criteria:

- Support COUNT, SUM, MIN, MAX, and AVG in production execution.
- Merge AVG from its rewritten sum and count components rather than averaging partition averages.
- Decode Gateway rewritten aggregate payloads and preserve SELECT VALUE versus aliased projection shapes.
- Emit one terminal aggregate result page after all required partitions are drained.
- Initially reject external continuation tokens explicitly.
- Cover empty inputs, mixed values, null, undefined, numeric overflow, aliases, multiple aggregates, and partition splits.
- Keep MAKE_LIST and MAKE_SET as explicit follow-up work pending cross-SDK parity confirmation.

### Use Case

Applications need server-compatible aggregate results over all partitions without manually fan-out querying and combining partial values.

### Alternatives

Application-side aggregation cannot reliably reproduce rewritten payload semantics, alias/value output shapes, diagnostics, split behavior, or Cosmos numeric handling.

### Additional Context

Implement under `sdk/cosmos/azure_data_cosmos_driver/src/driver/dataflow/`. This depends on the parent issue's rewritten-query application, aggregate result envelopes, buffering limits, response transformation, non-resumable continuation policy, and reusable dataflow tests. Cross-SDK references: .NET Aggregate pipeline stages, `SingleGroupAggregator`, accumulator files, and `AggregateQueryTests`; Java `AggregateDocumentQueryExecutionContext`, `SingleGroupAggregator`, and `aggregation/`.

Start in sdk/cosmos/azure_data_cosmos_driver/src/driver/dataflow/ and read the parent issue's rewritten-query application, aggregate result envelopes, buffering limits, response transformation, and non-resumable continuation policy. Compare the .NET Aggregate pipeline stages and SingleGroupAggregator and the Java AggregateDocumentQueryExecutionContext and aggregation/ references. Done means production COUNT, SUM, MIN, MAX, and AVG execution with the listed edge cases and reusable dataflow tests.

Written by the indexing model from the issue text.