multisign: decouple tx_list borrow lifetime from T's Transaction<'a> lifetime

[ckeshava]

Problem

transaction::multisign has a signature that conflates two lifetimes that
should be independent:

pub fn multisign<'a, T, F>(transaction: &mut T, tx_list: &'a Vec<T>) -> XRPLHelperResult<()>
where
    F: IntoEnumIterator + Serialize + Debug + PartialEq + 'a,
    T: Transaction<'a, F>,

'a is used both as:

1. The borrow lifetime of tx_list.
2. The data lifetime of T: Transaction<'a, F>.

Because they share one parameter, the compiler must pick a single 'a that
satisfies both. Once a caller's T lifetime is inferred as 'static — which
happens routinely when the transaction is built from string literals
(XRPAmount::from("20000000") → Cow::Borrowed<'static, str>) and used in
an async block that pushes the inferred lifetime to 'static — the borrow
of tx_list is also required to be 'static.

Observed symptom

Callers cannot pass a &local_vec to multisign from an async test that
also performs autofill: they have to Box::leak the Vec to obtain a
&'static Vec<T>. Concrete instance introduced by PR #298 (commit 6031ad2):

tests/transactions/multisign_payment.rs lines 81–90:
https://github.com/XRPLF/xrpl-rust/blob/6031ad2/tests/transactions/multisign_payment.rs#L81-L90

// multisign() merges the signer-signed copies into the master transaction.
// Box::leak the Vec because multisign's signature ties the borrow's lifetime
// to T's 'a parameter, which the async block infers as 'static when the
// Payment was built from string literals (e.g. XRPAmount::from("20000000")).
// Without 'static here the borrow checker rejects the call. The leak is
// intentional and bounded by the test runtime.
let signer_signed_copies: &'static Vec<_> =
    Box::leak(Box::new(vec![payment_signed_by_a, payment_signed_by_b]));
xrpl::transaction::multisign(&mut payment, signer_signed_copies)
    .expect("multisign combine");

Every external caller (not just tests) will hit this. Box::leak is not an
acceptable steady state for library users.

Proposed fixes

Option A — decouple borrow lifetime, accept a slice (minimal change)

pub fn multisign<'a, 'b, T, F>(transaction: &mut T, tx_list: &'b [T]) -> XRPLHelperResult<()>
where
    F: IntoEnumIterator + Serialize + Debug + PartialEq + 'a,
    T: Transaction<'a, F>,

'b is the short borrow lifetime; 'a remains Transaction's data lifetime.
The function body never stores tx_list past the call — it iterates once
and clones the first signer of each — so a short borrow is sufficient.

Bonus: switching &Vec<T> → &[T] is more idiomatic, and existing callers
passing &vec continue to work via Vec → slice coercion.

Option B — take an iterator, consuming the txs (more flexible)

pub fn multisign<'a, T, F, I>(transaction: &mut T, tx_list: I) -> XRPLHelperResult<()>
where
    I: IntoIterator<Item = T>,
    T: Transaction<'a, F> + Clone,
    F: IntoEnumIterator + Serialize + Debug + PartialEq + 'a,

Function consumes each signer-tx; no borrow at all. Callers pass
vec![a, b] directly. Trade-off: the call site now moves the Vec rather
than borrowing it, so callers can't reuse it afterward (rarely needed
here — the signer copies exist only to be merged).

Slightly larger refactor: tx_list: I is consumed, and the loop body
already does decoded_tx_signers.push(tx_signer.clone()) on the first
signer of each tx, so adapting from for tx in tx_list (which currently
borrows each element since tx_list: &Vec<T>) to for tx in tx_list
(now owning each element via IntoIterator) requires either
tx.get_common_fields() to remain valid on an owned T (it does), or
threading borrows differently. Either way, no borrow lifetime propagates
back to the caller.

Recommendation

Option A is the smallest backward-compatible change; option B is
structurally cleaner. Both eliminate the Box::leak requirement. I'd ship
A as a zero-risk fix and consider B alongside any future refactor of the
multisign ergonomics.

Acceptance

• Change the signature in src/transaction/multisign.rs to remove the
  borrow-lifetime / Transaction-lifetime coupling (option A or B).
• Update tests/transactions/multisign_payment.rs (PR feat: enforce 65% integration-test coverage and add tests #298, lines
  81–90) to drop the Box::leak.
• Verify the existing unit test in src/transaction/multisign.rs::tests
  still compiles.