Release eval: job for measuring request load

[Shaptic]

Parent: #706
Depends on: #707, #709

Summary

Use stellar-rpc-blaster in the release evaluation workflow to verify realistic read request load against a release-branch RPC binary.

Scope

• Add workflow setup for a pinned stellar-rpc-blaster version, binary, container, or source ref.
• Start stellar-rpc backfilled from an existing network and wait for healthy state.
• Define per-endpoint request profiles for read-heavy endpoints: getLatestLedger, getLedgers, getTransaction, getTransactions, getEvents, and getFeeStats.
• Run each endpoint profile for a configurable duration on the order of minutes and collect RPS, latency percentiles, error rate, timeout rate, and HTTP/RPC status breakdown.

Acceptance criteria

• The workflow includes a smoke profile for request load.
• Endpoint RPS targets are represented in a checked-in table with TBD placeholders until final values are provided.
• The job fails when measured RPS, latency, error-rate, or timeout criteria miss configured thresholds.
• Results are emitted per endpoint as structured JSON plus raw blaster output.
• Test data selection is deterministic enough for release-to-release comparison.

Notes

• GitHub-hosted runners may underrepresent production capacity; the goal is to catch large regressions, not certify maximum capacity.