Deploy the Sepolia testnet Bee as a light node + verify network availability

[crtahlin]

Part of #240. Depends on #SUB1.

Run a Sepolia testnet Bee as a light node on the staging VM, and prove it is actually connected to the network before trusting any upload.

Why a light node

The gateway buys batches via the chain client and stamps are signed off-chain, so staging only needs upload + pushsync. A light node (no reserve storage, no BZZ stake) is sufficient and far cheaper than a full node.

Scope

• Deploy Bee (container) for Sepolia testnet (chain id 11155111), full-node=false, swap enabled.
• Fund the Bee chequebook with Sepolia xDAI (faucet) and let it deploy the chequebook.
• Expose the Bee API on localhost only (gateway reads it); p2p port 1634 public.

Network-availability verification (must pass — #236)

• connectedPeers climbs to a healthy count (not stuck low).
• networkAvailability=Available (NOT Unavailable).
• Confirmed via the gateway debug proxy (/api/v1/debug/bee/topology, /status) and /health node section once the gateway is up (#SUB3).

Acceptance criteria

• Sepolia light Bee running and surviving restart.
• Chequebook funded + deployed on Sepolia.
• Peer count healthy + networkAvailability=Available.
• Bee API not exposed publicly; only p2p is.

[crtahlin]

Not done, and deliberately so. Closing this would misrepresent what is running; leaving it silent would too.

The plan was a Sepolia testnet Bee. What runs is a mainnet node:

swarm_connect-bee-dev-1  BEE_MAINNET=true

That followed from the decision to run one box for both environments, and from wanting staging to behave like production rather than differently. A testnet node would not have surfaced the issues staging actually caught this month — the shared-RPC rate limiting (#296), the postage amount floor (#258), the batch-status contradictions (#252) — all of which depend on real chain economics.

The cost is real and worth stating: staging spends real BZZ, which is why it currently sits at 0.0139 BZZ with an empty pool and cannot exercise any stamp flow.

Two ways forward, and this needs a decision rather than a default:

• Accept it, close this, and keep staging funded as an operating cost.
• Add a testnet node alongside, so cheap flows run free and only the ones needing mainnet economics spend.

Leaving open pending that call.