XDR view extractors: events, tx-hashes, tx-details, tx-pages from LedgerCloseMetaView #764

TL;DR

Implement four extractor functions that derive the data shapes RPC v2 needs from an xdr.LedgerCloseMetaView, using zero-copy aliasing into the source LCM. The perf benefits of zero-copy XDR views were validated on rpc-hack, so v2 will adopt them. Most of this code exists on rpc-hack already; the task is largely promoting it from bench-quality into a production package.

What to build

Four extractors over xdr.LedgerCloseMetaView:

• events — yields events-index payloads.
• tx-hashes — yields the txHash for each transaction in the LCM.
• tx-details by hash — given an LCM + target txhash, returns the transaction's envelope / result / meta.
• tx-pages — yields transactions in order with cursor support for the paginated getTransactions API.

All outputs zero-copy, aliasing into the source LCM buffer.

The stellar-rpc-side code lives in cmd/stellar-rpc/internal/fullhistory/views/ (or similar).

Determine the SDK vs stellar-rpc split

Part of this task: decide which pieces belong in go-stellar-sdk (alongside the view types) versus stellar-rpc. Rule of thumb — low-level XDR navigation primitives go to the SDK; RPC-specific data shapes (e.g., events.Payload, pagination cursors) stay here.

Source material on rpc-hack

Most of this code exists on rpc-hack already and can be cleaned up rather than written from scratch:

• events — cmd/stellar-rpc/internal/events/ingest_view.go (production-ready; originally in PR events,eventstore: concurrent index; migrate off membitmaps #756, will be scoped out of that PR and land via this task instead).
• tx-hashes / tx-details by hash — cmd/stellar-rpc/scripts/bench-fullhistory/tx_hash_helpers.go, extract_views.go, bench_cold_txhash.go (bench-quality; needs promotion).
• tx-pages — bench_cold_txpage.go / bench_hot_txpage.go for reference; reusability TBD.

Tests + benchmarks

• Per-extractor unit tests.
• Per-extractor go test -bench benchmarks so perf regressions are catchable.
• Differential correctness against the full-decode (UnmarshalBinary → derive) path on a checked-in fixture set covering recent ledgers, protocol-transition ledgers, and known edge cases (empty / sponsorship / large tx / V0/V1/V2 variants).
• Follow-up (separate issue): a harness that walks every ledger in history and asserts view-extraction matches full-decode. Not blocking the initial PR.

Acceptance

• Four extractors implemented with zero-copy outputs.
• SDK vs stellar-rpc split decided.
• Tests + benchmarks present for each.
• Fixture-set differential tests passing.
• Follow-up issue filed for the full-history sweep harness.