Refactor join handler

[cjen1-msft]

Refactor to pull up the various early-exit checks.

The primary effect is that this ensures that we have a single edge in the join process.
Specifically if a node is requesting to join, if it ever is successful, it will always be successful (see the check_node_exists).
Previously we had the seqno check above that, which would mean if you hit different nodes in the network you'd possibly get different responses.

The other nice effect is that it makes the function a simple series of do we early exit and then join.

[Copilot]

Pull request overview

Refactors the /node/join handler in the node RPC frontend to consolidate early-exit checks and ensure consistent “already joined” behavior (ie, once a join succeeds for a node identity, subsequent joins return the same result rather than failing due to ordering of validation checks).

Changes:

• Moves the “node already exists” check earlier so it short-circuits other validation (including the startup seqno gate).
• Simplifies the join flow into: validate target/service → return existing-node response if present → redirect if non-primary → validate startup seqno → add node with status based on service state.
• Replaces the is_taking_part_in_acking() usage with a direct NodeStatus::TRUSTED check.