lnwallet/chancloser: use valid delivery scripts in RBF closer tests

[ziggie1984]

Change Description

TestRbfChannelActiveTransitions shadows the package-level localAddr and
remoteAddr with twenty raw bytes each:

localAddr := lnwire.DeliveryAddress(bytes.Repeat([]byte{0x01}, 20))
remoteAddr := lnwire.DeliveryAddress(bytes.Repeat([]byte{0x02}, 20))

Neither is a well-formed delivery script — not a witness program, and not any
other template a co-op close is willing to pay to — so both are rejected by
validateShutdownScript.

Why it goes unnoticed today

validateShutdown is a chain of guards with early returns:

order	guard	error
1	thaw height not reached	ErrThawHeightNotReached
2	taproot shutdown nonce missing	ErrTaprootShutdownNonceMissing
3	delivery script invalid	ErrInvalidShutdownScript

Both subtests that feed the shadowed remoteAddr into a ShutdownReceived —
remote_initiated_thaw_height_close_fail and
remote_initiated_taproot_no_nonce_fail — trip an earlier guard on purpose, so
guard 3 is never reached. The one subtest that drives the whole chain,
remote_initiated_close_ok, was extracted in #11019 into
testRemoteInitiatedCloseOkNonTap / testRemoteInitiatedCloseOkTaproot, which
sit outside the shadow and therefore pick up the valid package-level scripts.

To be clear: nothing currently passes for the wrong reason. The harness
asserts on specific sentinel errors via errors.Is. But that holds only by
accident of where the failures happen to land, not by design.

Why it is worth fixing anyway

The cost is paid by whoever touches that guard chain next, and it is not paid
in a readable failure. Hoisting the delivery-script check above the thaw-height
check — a reasonable thing to want, since it is the cheaper check — does not
produce an assertion diff. The unexpected ErrInvalidShutdownScript reaches the
mock error reporter as an unmatched call and panics the package's test
binary:

panic: mock: Unexpected Method Call
  ReportError(*errors.errorString)
      0: &errors.errorString{s:"invalid shutdown script"}

The stack points into protofsm/state_machine.go and mock.go, not at the
fixture that is actually wrong.

Deleting the two shadowing locals with that same reorder still applied: the
whole package passes. So the shadow is the sole cause.

There is a second, quieter cost. local_initiated_close_ok passes the invalid
localAddr as localUpfrontAddr and asserts it lands in state unchallenged.
#11019 only added validation on the remote side; if anyone later makes that
symmetric, this test breaks for fixture reasons rather than subject reasons.

The fix

Drop the shadowing locals so those subtests use the valid package-level P2TR
scripts, and record the invariant where the scripts are declared. The general
principle: a negative test's fixture should be valid in every dimension except
the one under test, or it is not isolating what it claims to.

Where this came from

Found while backporting #11019 to v0.20.x. On that branch
remote_initiated_close_ok is still an inline subtest rather than an extracted
helper, so it sits inside the shadow and picks up the 20-byte blob. Once
#11019 makes delivery-script validation unconditional, that subtest fails with
invalid shutdown script. Master escapes this only because of the helper
extraction described above. The backport carries the same fixture fix; this PR
brings it to master, where it is latent rather than active.

Steps to Test

go test ./lnwallet/chancloser/... -race -count=1

To reproduce the latent failure this prevents, hoist the
validateRemoteDeliveryScript call to the top of validateShutdown in
lnwallet/chancloser/rbf_coop_transitions.go and run the package tests — on
master the binary panics; with this change it passes.

Pull Request Checklist

Testing

• Your PR passes all CI checks.
• Tests covering the positive and negative (error paths) are included.
• Bug fixes contain tests triggering the bug to prevent regressions.

Code Style and Documentation

• The change is not insubstantial. Typo fixes are not accepted to fight bot spam.
• The change obeys the Code Documentation and Commenting guidelines, and lines wrap at 80.
• Commits follow the Ideal Git Commit Structure.
• Any new logging statements use an appropriate subsystem and logging level.
• Any new lncli commands have appropriate tags in the comments for the rpc in the proto file.
• There is a change description in the release notes, or [skip ci] in the commit message for small changes.

Added under ## Testing in release-notes-0.22.0.md. This is a test-only
change with no user-visible effect, so no-changelog would also be reasonable
if maintainers prefer to keep that section for feature-level test work.

[github-actions]

🟢 PR Severity: LOW

gh pr view | 2 files | 15 lines changed

🟢 Low (2 files)

• docs/release-notes/release-notes-0.22.0.md - release notes documentation update
• lnwallet/chancloser/rbf_coop_test.go - test-only change (*_test.go)

Analysis

Both changed files fall into the LOW tier: one is a release-notes markdown update, and the other is a test-only file (*_test.go) under lnwallet/chancloser/. Per the classification rules, test files are categorized by their test-only nature rather than the package they live in, and are excluded from the file/line counts used for severity bumps. With only 2 files and 15 lines changed overall, no bump conditions apply.

---

_{To override, add a severity-override-{critical,high,medium,low} label.}