Add encoded length methods to transactions

[manuelmauro]

Summary

Add encoded_len() and payload_len() methods to compute transaction sizes without full encoding. This enables transaction size validation for DoS protection, matching reth/alloy's approach.

Motivation

Transaction pools need to enforce size limits for performance reasons and to prevent DoS attacks. Reth enforces a default maximum of 128 KB per transaction and validates transactions.

Alloy provides similar methods on transaction envelopes for this purpose.

[librelois]

Please add some unit tests, suggested scenarios:

1. encoded_len matches actual encoded bytes: for TransactionV0..V3, assert tx.encoded_len() == tx.encode().len().

2. payload_len matches payload encoding: assert tx.payload_len() == tx.encode_payload().len() for each tx variant.

3. Typed vs legacy size rule: assert typed tx (2930/1559/7702) has encoded_len() == payload_len() + 1, and legacy has encoded_len() == payload_len().

4. Unsigned message length methods: for Legacy/EIP2930/EIP1559/EIP7702 message structs, assert message.encoded_len() == rlp::encode(&message).len().

5. Size growth sanity: mutate fields with variable length (large input/data, longer access_list, larger authorization_list) and assert encoded_len increases.

6. Boundary-ish DoS shape case: construct a near-limit large tx payload (e.g., ~128KB data field) and verify length methods still exactly match full encoding length.

[manuelmauro]

Please add some unit tests, suggested scenarios:

1. `encoded_len` matches actual encoded bytes: for `TransactionV0..V3`, assert `tx.encoded_len() == tx.encode().len()`.

2. `payload_len` matches payload encoding: assert `tx.payload_len() == tx.encode_payload().len()` for each tx variant.

3. Typed vs legacy size rule: assert typed tx (`2930/1559/7702`) has `encoded_len() == payload_len() + 1`, and legacy has `encoded_len() == payload_len()`.

4. Unsigned message length methods: for `Legacy/EIP2930/EIP1559/EIP7702` message structs, assert `message.encoded_len() == rlp::encode(&message).len()`.

5. Size growth sanity: mutate fields with variable length (large `input/data`, longer `access_list`, larger `authorization_list`) and assert `encoded_len` increases.

6. Boundary-ish DoS shape case: construct a near-limit large tx payload (e.g., ~128KB data field) and verify length methods still exactly match full encoding length.

Addressed in 0d9c2e4

[librelois]

encoded_len()/payload_len() still perform full payload encoding and allocation, so they do not deliver the intended "size check without full encoding" DoS-hardening behavior.

Evidence:
src/enveloped.rs:48 uses self.encode_payload().len() for payload_len().
src/transaction/legacy.rs:288, src/transaction/eip2930.rs:250, src/transaction/eip1559.rs:125, src/transaction/eip7702.rs:289 each compute length via rlp::encode(self).len().
Impact: size validation remains allocation-heavy on untrusted inputs, which weakens the performance/DoS motivation.

Please implement true non-allocating length paths using RlpStream length accounting (begin_unbounded_list + append + finalize_unbounded_list) or explicit per-field RLP length math, and override payload_len() per transaction type.

[manuelmauro]

encoded_len()/payload_len() still perform full payload encoding and allocation, so they do not deliver the intended "size check without full encoding" DoS-hardening behavior.

Evidence: src/enveloped.rs:48 uses self.encode_payload().len() for payload_len(). src/transaction/legacy.rs:288, src/transaction/eip2930.rs:250, src/transaction/eip1559.rs:125, src/transaction/eip7702.rs:289 each compute length via rlp::encode(self).len(). Impact: size validation remains allocation-heavy on untrusted inputs, which weakens the performance/DoS motivation.

Please implement true non-allocating length paths using RlpStream length accounting (begin_unbounded_list + append + finalize_unbounded_list) or explicit per-field RLP length math, and override payload_len() per transaction type.

Addressed in ee70e83

[librelois]

length_of_length is capped at 4, so it undercounts RLP headers when payload length exceeds 0xFFFF_FFFF bytes (4 GiB). That propagates into list/string length math and can make encoded_len/rlp_len incorrect for very large payloads.
File: src/transaction/rlp_len.rs (line 28), src/transaction/rlp_len.rs (line 131)

Suggestions:

1. Fix length_of_length to compute bytes needed for arbitrary usize (not capped at 4).
2. Add unit tests around boundary values (0xFF, 0xFFFF, 0xFFFFFF, 0xFFFFFFFF, 0x1_0000_0000 on 64-bit) for rlp_list_header_len/byte-string length helpers.

[manuelmauro]

length_of_length is capped at 4, so it undercounts RLP headers when payload length exceeds 0xFFFF_FFFF bytes (4 GiB). That propagates into list/string length math and can make encoded_len/rlp_len incorrect for very large payloads. File: src/transaction/rlp_len.rs (line 28), src/transaction/rlp_len.rs (line 131)

Suggestions:

1. Fix length_of_length to compute bytes needed for arbitrary usize (not capped at 4).

2. Add unit tests around boundary values (0xFF, 0xFFFF, 0xFFFFFF, 0xFFFFFFFF, 0x1_0000_0000 on 64-bit) for `rlp_list_header_len`/byte-string length helpers.

Addressed in d813d5d