fix: optimize Firo node config for memory-constrained VM

[bernd2022]

Summary

• Optimize Firo node configuration to prevent swap thrashing on 3.8GB Azure VM running 5 blockchain containers
• Node was freezing due to memory pressure causing lock starvation (peak usage 2.27GB with old config)
• After optimization: stable at ~1.2GB, startup in ~4 minutes, node healthy

Changes

firo.conf

• dbcache: 512 → 256 MB (reduced UTXO cache)
• maxmempool: 300 → 100 MB (limit mempool memory)
• mempoolexpiry: 72 → 24 hours (faster cleanup)
• maxorphantx: 100 → 10 (reduce cs_main lock hold time during Spark tx validation)
• rpcworkqueue: 32 → 64 (more buffer before HTTP 500 rejection)
• Per-connection send/receive buffers reduced

docker-compose-firo.yml

• Add Docker memory limit (2048M) and reservation (1024M)
• Healthcheck: getblockcount instead of getblockchaininfo (lighter RPC call)
• Healthcheck interval: 30s → 120s (prevent RPC thread exhaustion)
• Healthcheck timeout: 60s → 10s, retries: 10 → 3

Test plan

• Deployed to dfx-out-dev, node running stable and healthy
• Full sync achieved (progress=1.000000)
• Memory usage stable at ~1.2GB / 2GB limit (58%)
• Monitor stability over next days