fix(core): validate StorageProof structure in CheckTx

[raymondjacobson]

Summary

`CheckTx` previously accepted any transaction that parsed as valid protobuf — no per-type validation. This meant a malformed `StorageProof` (e.g. with empty `ProverAddresses`) could enter the mempool and get gossiped across the network, even though it would later be rejected at `ProcessProposal` (#315).

This PR adds a `checkTxStructural` step to `CheckTx` that performs lightweight, context-free validation before a transaction enters the mempool. For `StorageProof` txs it checks:

• `ProverAddresses` is non-empty
• `Address` (prover comet address) is set
• `Height` is non-zero

These checks are purely structural — no DB lookups, no block height needed — so they add negligible latency to `CheckTx`.

The `checkTxStructural` method uses a type switch and is easy to extend with structural checks for other tx types in the future.

Defense in depth (full picture after #313, #315, this PR)

Layer	Guard	Prevents
Submission (`sendPoSChallengeToStorage`)	Early-return on empty proverAddresses (#315)	Bad tx from being created locally
CheckTx (`checkTxStructural`)	Reject empty ProverAddresses (this PR)	Bad tx from entering mempool / gossip
ProcessProposal (`isValidStorageProofTx`)	Reject empty ProverAddresses (#315)	Bad tx from being included in a block
FinalizeBlock (`finalizeStorageProof`)	Coerce nil → `[]string{}` (#313)	NULL constraint violation if all else fails

Test plan

• CI passes
• Submit a `StorageProof` tx with empty `ProverAddresses` via `SendTransaction` RPC — should get rejected with code 1 (not enter mempool)
• Normal `StorageProof` txs with valid fields still pass `CheckTx`

🤖 Generated with Claude Code