fix(retry): fix sporadic "no request forwarding" errors when fetching discovery members through a VIP endpoint

[perfectra1n]

Fixes sporadic "no request forwarding" errors when fetching discovery members through a VIP endpoint.

When the TUI queries discovery members via the VIP, the request may route to a control plane node that can't forward it, causing the discovery_members to be cleared. Therefore falling back to an empty node list, and the TUI showing empty/stale cluster data.

Retry with fallback nodes, first tries VIP (2 retries), then queries each control plane node directly. I also set up to preserve cached data and don't clear discovery_members on error. Then finally, tertiary fallback where we use versions data as node source if discovery AND etcd both fail.

File	Change
talos-rs/src/talosctl.rs	Add get_discovery_members_for_node_async() and update retry function to accept fallback node IPs
talos-rs/src/lib.rs	Export updated function
talos-pilot-tui/src/components/cluster.rs	Pass etcd member IPs as fallback, add tertiary fallback using versions, preserve data on error
talos-pilot-tui/src/components/lifecycle.rs	Pass version node IPs as fallback, preserve data on error

Let me know if you have any questions or any suggestions :)

[perfectra1n]

I ran into this issue when having my Talos config set up like:

context: mycluster
contexts:
    mycluster:
        endpoints:
            - 192.168.9.11
            - 192.168.9.12
            - 192.168.9.13
            - 192.168.9.21
            - 192.168.9.22
            - 192.168.9.23
        nodes:
            - 192.168.9.11
            - 192.168.9.12
            - 192.168.9.13
            - 192.168.9.21
            - 192.168.9.22
            - 192.168.9.23

Where it would just randomly fail at startup...