Test that proofsForTransactions() does not modify the UTXO set tree along a growing chain

[Ergologica]

Addresses #1614.

What was already covered, and what was missing. UtxoStateSpecification already checks that a proofsForTransactions() call does not change the state digest on a fresh state ("proofsForTransactions() does not change state digest") and that it is deterministic there ("proofsForTransactions() to be deterministic"). No test checked the same invariants while a chain grows - a tree with real history, with proofs generated between block applications, which is how the node actually uses it. I asked about this reading of the scope in an issue comment on 2026-08-11 and heard no objection.

The new property, proofsForTransactions() does not modify the UTXO set tree along a growing chain, builds a chain of valid full blocks the same way the neighbouring concurrency test does, and at every height asserts:

1. the tree root is unchanged after generating proofs (rootDigest before == after);
2. a second call returns the identical proof digest and state digest, and still leaves the root untouched;
3. the block assembled from those proofs (stateRoot = adDigest, ADProofsRoot = ADProofs.proofDigest(...)) is accepted by applyModifier, and the state lands exactly on the promised digest.

On the numbers in the issue: the literal 500 blocks x 500 transactions would take hours in CI, so the chain here is 20 blocks with the standard generator transactions - the invariants are the same at any scale, and the length is a one-line constant. Happy to raise it, or to add an ignored long-run variant with the literal numbers, if preferred.

Verified the test bites: injecting a mutation into proofsForTransactions (staging an extra Insert on the persistent prover and committing it) makes the property fail on the root check; on clean master it is green, and the full UtxoStateSpecification suite passes locally (22/22).

Note: the issue shows a years-old assignee with no linked work; if that work is still planned I am happy to defer.

[Ergologica]

CI note: the red job is Run it node tests. Both failures are Docker multi-node convergence timeouts, in suites this PR does not touch:

• DeepRollBackSpec ("Deep rollback handling") - TimeoutException: Nodes did not converge to the same best full block at height >= 255
• ForkResolutionSpec ("Fork resolution after isolated mining") - TimeoutException: Futures timed out after [15 minutes]

This PR only adds one property to UtxoStateSpecification, which runs in Run node tests - green on this run, as are the wallet and ergo-core jobs; 11 of 13 integration tests passed too, including UtxoStateNodesSyncSpec. DeepRollBackSpec timeouts have hit other PRs as well (see #2465 / #2469, "Retrigger CI after DeepRollBackSpec timeout"). Happy to rebase to retrigger if a green re-run is preferred.