Add async iterable support for paginating query, scan, and collection results

[anatolzak]

Summary

The result of .go() on query, scan, and collection operations should support for await...of iteration, yielding one page of results at a time. This enables streaming-style pagination without manual cursor management.

Motivation

Currently, consuming all pages of a query requires either:

1. Setting { pages: "all" } and waiting for all pages to accumulate into a single response, or
2. Manually looping with cursors:

let cursor = null;
do {
  const results = await entity.query.myIndex({ pk }).go({ cursor });
  // process results.data
  cursor = results.cursor;
} while (cursor !== null);

Neither approach gives you per-page control with a clean API. The cursor loop is verbose, and { pages: "all" } forces you to wait for all data before processing any of it.

Proposed API

The result of .go() becomes dual-consumable — it works as a Promise (existing behavior, unchanged) and as an AsyncIterable (new):

// Existing behavior — unchanged
const result = await entity.query.myIndex({ pk }).go({ pages: "all" });
console.log(result.data);

// New — iterate page by page
for await (const page of entity.query.myIndex({ pk }).go({ pages: "all" })) {
  console.log(page.data);   // items for this page
  console.log(page.cursor);  // cursor for this page
}

// Works with scans
for await (const page of entity.scan.go({ pages: "all" })) {
  console.log(page.data);
}

// Works with collections
for await (const page of service.collections.myCollection({ id }).go({ pages: "all" })) {
  console.log(page.data.entityA);
  console.log(page.data.entityB);
}

// Early termination
for await (const page of entity.scan.go({ pages: "all" })) {
  items.push(...page.data);
  if (items.length >= 100) break;
}

Key behaviors

• No breaking changes — await .go() continues to work exactly as before
• Lazy — the underlying query is not executed until .then() or for await is called
• Per-page results — each yielded value has { data, cursor } (and unprocessed when hydrate is used)
• All execution options respected — limit, pages, count, data, attributes, etc. all work with iteration
• Early break supported — breaking out of the loop stops further DynamoDB requests