Fix: use separate env var for ClickHouse native port

[mpont91]

The analytics service talks to ClickHouse over two protocols, but both read CLICKHOUSE_PORT:

• clickhouse_connect over HTTP (8123)
• clickhouse_driver over native TCP (9000) — in nav_calculator.py, scheduler.py and run_all.py

The Dockerfile pins CLICKHOUSE_PORT=8123, so the native clients hit the HTTP port and fail:

nav_calculator - ERROR - Failed to calculate NAV for PSI-10: Code: 102.
Unexpected packet from server clickhouse:8123 (expected Hello or Exception...)
...same for all 7 funds -> "Calculated NAV for 0 funds"

No single value works for both, so this adds CLICKHOUSE_NATIVE_PORT (default 9000) for the three native call sites. CLICKHOUSE_PORT keeps its HTTP meaning.

Verified locally: NAV goes from 0 to 7 funds, no errors. Also applies to deploy/docker-compose.prod.yaml, which had the same setting.