Cardano Updates

Adds a Rosetta-inspired, read-only test suite at tests/ for any GraphQL
implementation of the cardano-graphql schema. Six phases shipped in one go:

  Phase 1 — sanity, golden functional, negative tests
  Phase 2 — Locust load harness + p95<500ms / err<1% baseline gate
  Phase 3 — multi-instance comparison (skipped without --compare-url)
  Phase 4 — regression goldens seeded from cardano-foundation issues
            (#951, #953, #978/#980, #979)
  Phase 5 — schema introspection diff against committed snapshot
  Phase 6 — CI workflows: PR gate, nightly, weekly

Coverage:
  * 52 / 52 top-level Query roots (100%)
  * 428 / 440 data-type fields (97.3%) across 61/61 types touched
  * 59 mainnet goldens captured at anchor block 13228000
  * 8 preprod goldens scaffolded (auto-skip on mainnet runs)
  * 8 negative tests, 7 schema tests (1 integration + 6 offline unit)

Determinism:
  * All goldens pinned to anchor_block 13228000 in config/networks.yaml
  * Address buckets: light / medium / heavy with separate goldens + perf tasks
  * --update-golden record-then-review lifecycle; forbidden when CI=true
  * Volatile fields masked via dot.notation[*] ignore_paths DSL

Tooling:
  * uv-managed, Python 3.12+
  * generate_report.py: self-contained HTML report with coverage panel,
    per-test timing table, and query corpus appendix
  * utils/coverage.py: graphql-core-based coverage report (top-level + field)
  * utils/markers.py: lists 52 auto-applied root markers + cip26/cip68
  * Dynamic markers — pytest -m assets / -m cip26 / -m cip68 / -m <root>
    target only goldens whose query touches that root, computed at collection

Reusable patterns lifted from cardano-rosetta-java/tests/data-endpoints:
  * pytest_runtest_makereport Allure attachment hook
  * Recording GraphQL client proxy
  * YAML network test data layout
  * Deep-diff with path-based ignores

Documentation:
  * tests/README.md            — top-level reference
  * tests/QUICKSTART.md         — install + run in 5 minutes
  * tests/TESTING_GUIDE.md      — skim-friendly reference for all test types
  * tests/CONTRIBUTING.md       — how to add tests, refresh anchor, debug
  * tests/{performance,multi_instance,schema,negative}/README.md
  * tests/queries/{coverage,regression}/README.md
  * .claude/skills/test-cardano-graphql.md — natural-language Claude skill

CI:
  * .github/workflows/graphql_tests.yml — PR gate (markers=pr), nightly (full
    functional + negative + schema), weekly (adds perf smoke + baseline gate),
    workflow_dispatch with optional URL override and run_perf toggle.
  * Every job exports CI=true to block --update-golden / --update-schema.