Fix missing PaymentSent::fee_paid_msat

[valentinewallace]

If an outbound payment was abandoned with htlcs in-flight and later claimed, we would previously have the PaymentSent::fee_paid_msat be set to None. This contradicted some docs on the event that stated the field would always be Some after 0.0.103.

Fixes #4639

[ldk-reviews-bot]

I've assigned @tankyleo as a reviewer!
I'll wait for their review and will help manage the review process.
Once they submit their review, I'll check if a second reviewer would be helpful.

[ldk-claude-review-bot]

nit: This wording slightly overstates the scope. For payments that stayed in Retryable state (i.e., were never abandoned while in-flight), fee_paid_msat has been Some since 0.0.103. The None gap between 0.0.103 and 0.4 only applies to payments that were abandoned with HTLCs in-flight and then unexpectedly claimed.

Something like "May be None for payments abandoned on LDK versions prior to 0.4, or initiated prior to 0.0.103." would be more precise. Not blocking — the current "May be" phrasing is technically correct.

[ldk-claude-review-bot]

Review Summary

The fix is correct. The core logic — preserving pending_fee_msat during the Retryable → Abandoned transition and keeping it in sync as failed paths are removed — is sound. I verified the key invariants:

1. In mark_abandoned(), get_pending_fee_msat() is called while still in Retryable state (before the transition), so the fee is correctly captured.
2. In handle_pay_route_err(), remove() is called before auto-abandonment (mark_abandoned), so the fee is decremented while still in Retryable state for the auto-abandon case. For the manual-abandon case (abandon_early=true), the new Abandoned branch in remove() handles it.
3. send_payment_success() reads get_pending_fee_msat() before calling mark_fulfilled(), so the fee is captured before transitioning to Fulfilled (which doesn't track fees).
4. The serialization with TLV field 5 (odd = optional) is backward-compatible — old nodes without this field will deserialize pending_fee_msat as None.

The test assertion change from Some(None) to Some(Some(2_000)) in the existing test_inconsistent_mpp_params is also correct — that test's payment gets auto-abandoned when the mismatched MPP path fails (since retry_strategy is None), so it was hitting the same Abandoned-state bug even without explicit abandonment.

Inline comments posted:

• lightning/src/events/mod.rs:1204 — doc precision: the None gap between 0.0.103 and 0.4 only applies to abandoned-then-claimed payments, not all payments.

No bugs found.