docs: Add reliable transaction pagination guidance and example

[arunimshukla]

Summary

• Correct the README's claim that 100 entities is a hard per-query limit; 100 is the default collection size.
• Replace the offset-only example with an ID cursor, a fixed finalized block hash, and per-page deployment/snapshot checks.
• Add a dependency-free, read-only Node.js example scoped to one wallet's V2/V3 transaction history.
• Add local tests for cursor advancement, tied timestamps, short and empty pages, HTTP/GraphQL errors, snapshot drift, and CLI failure handling.
• Check that the README query matches the tested example and that selected fields exist in both current protocol schemas.

Why

The existing pagination paragraph confuses the default page size with a hard limit. It also provides no guidance for stable historical exports or for distinguishing failed/partial responses from complete results.

The updated guidance follows The Graph's querying best practices and GraphQL API documentation.

Related to #155 as query-troubleshooting guidance. This does not reproduce or diagnose that issue's position-query inconsistency, claim pagination caused it, or attempt to close it.

Scope

• README and standalone example/test files only.
• No schema, mapping, contract, deployment, dependency, or package-script changes.
• No affected indexer deployment or new test deployment. The example issues read-only queries against an endpoint supplied by the caller.
• Cursor exhaustion is not independent proof that all on-chain events were indexed correctly; the README makes that boundary explicit.

Validation

node --test examples/paginate-user-transactions.test.mjs

• 17 tests passed on Node.js 24.19.0.
• Verified the original README against upstream blob 5cc46fb064e55a488043c3a06e8e512425bedad1.
• Reviewed query fields against schemas/v2.schema.graphql and schemas/v3.schema.graphql, and ID generation against src/utils/id-generation.ts.
• Verified one contribution commit containing exactly the three intended files.
• Tests use mocked responses. No live endpoint execution, deployed GraphQL schema validation, or subgraph build/deployment is claimed.