fix(framework): handle shielded admission and validate note logs

[Federico2014]

What does this PR do?

Propagate local transaction admission results to RPC and P2P callers, and add defensive length validation to Shielded TRC20 note scanning.

• When Manager.pushTransaction returns false, RPC returns result=false with SERVER_BUSY and skips broadcast. The response message and logs describe local admission rejection without assuming a specific Manager rejection reason.
• Return the admission result from TronNetDelegate.pushTransaction; the P2P handler drops rejected transactions without rebroadcasting or penalizing the peer.
• After contract-address and topic identification, require exactly 800 bytes for MintNewLeaf, TransferNewLeaf, and BurnNewLeaf logs in IVK and OVK scanning. Use consistent length conditions and preserve the separate TokenBurn validation.
• Re-enable Shielded TRC20 builder tests and correct their ABI encoding, value balances, outgoing viewing keys, change outputs, and frontier updates.

Why are these changes required?

A full native shielded pending pool makes Manager.pushTransaction return false. Previously, RPC and P2P callers ignored that result and continued broadcasting; RPC could report success even though the transaction had not been admitted locally.

Note scanning already identifies supported events by contract address and topics. The explicit length check is defensive hardening that makes the expected ABI layout clear: four 32-byte fields plus bytes32[21] ciphertext data total 800 bytes, including 12 bytes of ciphertext padding.

This PR has been tested by:

• Local targeted tests passed: WalletMockTest.testBroadcastTransactionShieldedPendingPoolFull, ShieldedTRC20BuilderTest.testScanShieldedTRC20NotesSkipsInvalidLogLengths, and all four NoteEncDecryTest.testGetNoteTxFromLogListByOvk* tests.
• Admission tests verify SERVER_BUSY, retained txid cache entries, duplicate rejection on retry, and skipped broadcast. Existing delegate and P2P regression tests cover result propagation and no peer penalty on local rejection.
• Note scanning tests cover lengths 0, 788, 799, 800, 801, and 832 for all three note events through IVK and OVK. The TokenBurn tests cover successful decoding, short records, missing nullifiers, and cursor pairing.
• Local :framework:checkstyleMain passed.
• Manual data verification: sampled mainnet mint, transfer, and burn receipts from contract TQEuSEVRk1GtfExm5q9T8a1w84GvgQJ13V; their note logs were each 800 bytes. No live transaction submission was performed.

Follow up

Txid cache cleanup after admission rejection or validation exceptions, concurrent reservation ownership, and retry control will be handled in a separate PR. This PR preserves the existing cache lifecycle: with caching enabled, rejected transactions retain their txid entries, so retries can return DUP_TRANSACTION_ERROR even after pending capacity recovers.

Rejection metrics and delayed retry for re-push transactions can be considered separately.

Extra details

Admission handling applies to native ShieldedTransferContract transactions when that feature is enabled; log validation applies to Shielded TRC20 contract events.

RPC rejection now uses the existing SERVER_BUSY code instead of incorrectly reporting success. The Java return type of TronNetDelegate.pushTransaction changes from void to boolean: source callers may ignore the result, overriding subclasses must adapt, and previously compiled Java consumers must be recompiled. HTTP/gRPC schemas are unaffected by that Java signature change.

Re-push behavior is unchanged. No database schema, consensus, protocol, or configuration changes are included.