Cache resource service accessors

[terjekv]

Why

Every access to a fixed service property such as client.classes, client.groups, or client.tasks currently allocates a new wrapper around the same client. Normal workflows access these properties repeatedly, creating short-lived objects without changing behavior.

What changed

• lazily cache each fixed resource-service accessor on Client and AsyncClient
• preserve sync/async parity and stable service identity per client
• keep class selectors and class-scoped object services on demand through client.classes.by_id(class_id).objects
• add regression coverage for laziness, reuse, per-client isolation, and class-scoped factory behavior
• document service lifetimes and update the unreleased changelog

Verification

• uv sync --extra dev
• uv run ruff format --check .
• uv run ruff check .
• uv run mypy
• uv run bandit -q -r src scripts
• uv run zizmor .
• uv run pytest --cov — 327 passed, 99.05% coverage
• uv run mkdocs build --strict
• uv build

The pinned Hubuum v0.0.3 e2e workflow in CI remains the live compatibility authority for this PR.