fix(config): wire K8s env vars into HiveStack and API server

[cursor]

Bug and impact

K8s/Helm deployments set HIVE_VALIDATE_INPUTS, HIVE_RATE_LIMIT, HIVE_MAX_NODES, HIVE_MAX_CONTENT_BYTES, and HIVE_DEFAULT_TTL_S via ConfigMap/deployment env vars, but scripts/hive_api_server.py created a bare HiveStack() that never called HiveConfig.from_env(). Enterprise controls were silently ignored at runtime — input validation, rate limiting, memory caps, and content-size limits had no effect in the REST API path.

Concrete trigger: Deploy via deploy/k8s/ with HIVE_VALIDATE_INPUTS=true and HIVE_RATE_LIMIT=1000, then POST an oversized agent state to /route. Before this fix, validation and rate limiting never ran despite operator configuration.

Root cause

HiveConfig was implemented and documented for env-based configuration, but:

• hive_api_server.py did not load config from the environment
• HiveStack did not wire max_memory_nodes, max_content_bytes, or rate_limit from HiveConfig into RustBrain / RateLimiter

Fix

• Load HiveConfig.from_env() in the REST API server
• Pass max_nodes from config.max_memory_nodes to RustBrain
• Auto-create RateLimiter when HIVE_RATE_LIMIT > 0
• Honor max_content_bytes from config when not explicitly overridden
• Accept HIVE_MAX_NODES alias (used in K8s manifests)
• Fix AsyncHiveStack.remember() silently dropping trust/tags

Validation

• Added regression tests in tests/test_enterprise_config.py, tests/test_hive_api_server.py, and tests/test_async_stack.py
• Full suite: 184 passed, 6 skipped