feat: load epoch schedule from genesis.bin for warmup-aware epoch math

[Mctursh]

Loads the cluster's epoch schedule from a genesis.bin file so warmup-aware epoch math (getInflationReward) uses the real network schedule instead of a hardcoded constant.

Per review, this moved from storing the schedule in a ClickHouse table to reading it from genesis.bin, the canonical per-network source, so there's no full schema for one immutable row.

How it works

• --genesis-path / GENESIS_PATH (optional): path to the cluster's genesis.bin; the RPC reads the EpochSchedule from it once at startup into AppState.
• Unset: falls back to the no-warmup schedule (432000 slots), which is what mainnet and devnet return. Testnet runs a warmup schedule, so set --genesis-path there.
• Set but unreadable/unparseable: the server fails to start, instead of silently using the wrong schedule.

What changed

• Removed the epoch_schedule ClickHouse table (local/cluster/replicated DDL + the k8s apply step) and the ingestor-side getEpochSchedule fetch.
• getInflationReward reads the schedule from AppState (genesis or the fallback); its query path is otherwise unchanged.
• Genesis loading sits in its own module so the superbank-verify PoH work can reuse it.

Testing
Unit tests cover the genesis decode (round-trip + malformed + file-not-found) and the epoch-bounds math (mainnet-matches-old vs warmup-differs).

getEpochSchedule (#36) still returns a hardcoded schedule; wiring that to this shared schedule is the follow-up, along with getEpochInfo (#30) once it lands.

Closes #47.

[linuskendall]

Instead of storing this in the databases, couldn't this be loaded from a genesis.bin file? If you want to have epoch schedule calculations you can add genesis.bin to the config file of superbank?

Otherwise we add a full schema for just a single row that never changes.

[notwedtm]

Agreed. genesis.bin would provide us a more SVM generic interface that can support a wider range of chains/networks.

It's also going to be needed for the superbank-verify PoH workflows, so will serve multiple purposes if it's included.

[Mctursh]

Makes sense, agreed with both of you. I'll rework the PR to load the schedule from a genesis.bin path in the config instead of the ClickHouse table.

[Mctursh]

Pushed the rework. The epoch_schedule ClickHouse table and the ingestor fetch are gone. The RPC now loads the schedule from a genesis.bin (``--genesis-path/GENESIS_PATH`) at startup. No path set falls back to the standard no-warmup schedule, which is what mainnet and devnet actually return (checked against live RPC); testnet runs a warmup schedule, so point `--genesis-path` at its genesis there. A set-but-unreadable path fails boot instead of silently guessing.

Kept the genesis loading in its own module so it's easy to reuse for the superbank-verify PoH work later. getInflationReward reads the schedule from state the same way, just populated from genesis now.

[Mctursh]

GoReleaser hit File exists (os error 17) after every crate built. It passed on the parent commit 919cb87, and the only change since is a two-line doc edit, so it's not from this PR.