fix: check node-tx timelock exhaustion when refund-tx is healthy in checkRenewLeaves

[vwinee21]

Problem

Fixes #127

checkRenewLeaves gates the entire renewal classification on the refund-tx sequence. When refundSequence does not need renewal, the node is immediately pushed to validNodes — without ever checking whether nodeSequence has decayed to ≤ TimeLockInterval.

Result: a leaf whose node-tx timelock is exhausted but whose refund-tx timelock is still healthy is classified as Valid and selected for transfers. The SSP then attempts to construct the next CPFP refund and ValidateSequence rejects with:

current timelock 100 (rounded from 100) in CPFP refund transaction is too small
to subtract TimeLockInterval 100 without reaching zero

The leaf is structurally unspendable but checkRenewLeaves returns 0 renewals on every preflight — no ceremony attempted, no warn logs, leaf stays stuck indefinitely.

Fix

In the else branch (refund-tx healthy), add an explicit check for node-tx exhaustion before classifying the node as valid:

} else {
  if (doesTxnNeedRenewed(nodeSequence)) {
    nodesToRenewNodeTxn.push(node);
    nodeIds.push(node.id);
  } else {
    validNodes.push(node);
  }
}

This mirrors the existing logic inside the if (doesTxnNeedRenewed(refundSequence)) branch and ensures node-tx exhaustion is caught regardless of refund-tx health.