This commit doesn't contain any logical change, we just move code to
align with the FSM flow. It makes it easier to follow the progress of
the state machine to always scroll down when advancing states.

We refactor `NodeRelay.scala` to re-order some steps, without making
meaningful functional changes. The steps are:

1. Fully receive the incoming payment
2. Resolve the next node (unwrap blinded paths if needed)
3. Wake-up the next node if necessary (mobile wallet)
4. Relay outgoing payment

Note that we introduce a wake-up step, that will be enriched in future
commits and can be extended to include mobile notifications. The file
is now also easier to follow, as steps are done linearly by simply
scrolling down.

We allow relaying data to contain a wallet `node_id` instead of an scid.
When that's the case, we start by waking up that wallet node before we
try relaying onion messages or payments.

Create a dedicated `peer-wake-up` configuration section.
This can be enriched with mobile notification sub-sections.

It turns out that we can keep `requestedShortChannelId_opt` and
`walletNodeId_opt` as fields of this actor instead of resolving
them once and forwarding the value in every function.

This field is used in feature branches, so we include it to minimize
the diff.

This commit doesn't contain any changes apart from using a private class
to factor common fields of the `PeerReadyNotifier`.

- clean-up `SomePeerConnected` / `SomePeerDisconnected`
- include supervision directly inside the actor itself

It can be useful to track pending `PeerReadyNotifier` instances to avoid
performing duplicate actions when multiple `PeerReadyNotifier` spawn for
the same peer (e.g. sending a mobile notification).

When a `PeerReadyNotifier` actor is started, it registers itself into a
singleton `PeerReadyManager`, which tells it whether there are other
pending attempts for the same peer.

This way we can have `isFirstAttempt` as a class field, which makes
sense since this value won't change.

This makes testing easier. We also do small refactorings in the relay
actors without any behavior changes.