fix(observe): unify daemon logs through single channel

[mccormickt]

This PR unifies auraed's logging patterns through a single LogChannel, leveraging tracing_subscriber::MakeWriter and a new BroadcastWriter struct to properly broadcast log messages to all subscribers.

Before this change, aer observe get-aurae-daemon-log-stream would never display logs.

Now:

» aer observe get-aurae-daemon-log-stream
Streaming
GetAuraeDaemonLogStreamResponse {
    item: Some(
        LogItem {
            channel: "auraed",
            line: "{\"timestamp\":\"2026-09-12T12:53:56.653520Z\",\"level\":\"INFO\",\"fields\":{\"message\":\"Nested auraed running with host pid 3858842\"},\"target\":\"auraed::cells::cell_service::cells::nested_auraed::nested_auraed\",\"span\":{\"request\":\"ValidatedCellServiceAllocateRequest { cell: ValidatedCell { name: CellName(\\\"test\\\"), cpu: Some(ValidatedCpuController { weight: None, max: None, period: None }), cpuset: Some(ValidatedCpusetController { cpus: None, mems: None }), memory: Some(ValidatedMemoryController { min: None, low: None, high: None, max: None }), isolate_process: false, isolate_network: false } }\",\"name\":\"allocate\"},\"spans\":[{\"request\":\"ValidatedCellServiceAllocateRequest { cell: ValidatedCell { name: CellName(\\\"test\\\"), cpu: Some(ValidatedCpuController { weight: None, max: None, period: None }), cpuset: Some(ValidatedCpusetController { cpus: None, mems: None }), memory: Some(ValidatedMemoryController { min: None, low: None, high: None, max: None }), isolate_process: false, isolate_network: false } }\",\"name\":\"allocate\"}]}",
            timestamp: 1789217636,
        },
    ),
}
GetAuraeDaemonLogStreamResponse {
    item: Some(
        LogItem {
            channel: "auraed",
            line: "{\"timestamp\":\"2026-09-12T12:53:56.664483Z\",\"level\":\"INFO\",\"fields\":{\"message\":\"Attach nested Auraed pid 3858842 to cgroup test\"},\"target\":\"auraed::cells::cell_service::cells::cell\",\"span\":{\"request\":\"ValidatedCellServiceAllocateRequest { cell: ValidatedCell { name: CellName(\\\"test\\\"), cpu: Some(ValidatedCpuController { weight: None, max: None, period: None }), cpuset: Some(ValidatedCpusetController { cpus: None, mems: None }), memory: Some(ValidatedMemoryController { min: None, low: None, high: None, max: None }), isolate_process: false, isolate_network: false } }\",\"name\":\"allocate\"},\"spans\":[{\"request\":\"ValidatedCellServiceAllocateRequest { cell: ValidatedCell { name: CellName(\\\"test\\\"), cpu: Some(ValidatedCpuController { weight: None, max: None, period: None }), cpuset: Some(ValidatedCpusetController { cpus: None, mems: None }), memory: Some(ValidatedMemoryController { min: None, low: None, high: None, max: None }), isolate_process: false, isolate_network: false } }\",\"name\":\"allocate\"}]}",
            timestamp: 1789217636,
        },
    ),
}
GetAuraeDaemonLogStreamResponse {
    item: Some(
        LogItem {
            channel: "auraed",
            line: "{\"timestamp\":\"2026-09-12T13:06:31.753030Z\",\"level\":\"INFO\",\"fields\":{\"message\":\"CellService: free() cell_name=CellName(\\\"test\\\")\"},\"target\":\"auraed::cells::cell_service::cell_service\",\"span\":{\"request\":\"ValidatedCellServiceFreeRequest { cell_name: CellName(\\\"test\\\") }\",\"name\":\"free\"},\"spans\":[{\"request\":\"Request { metadata: MetadataMap { headers: {\\\"te\\\": \\\"trailers\\\", \\\"content-type\\\": \\\"application/grpc\\\", \\\"user-agent\\\": \\\"tonic/0.14.2\\\"} }, message: CellServiceFreeRequest { cell_name: \\\"test\\\" }, extensions: Extensions }\",\"name\":\"free\"},{\"request\":\"ValidatedCellServiceFreeRequest { cell_name: CellName(\\\"test\\\") }\",\"name\":\"free\"}]}",
            timestamp: 1789218391,
        },
    ),
}

We also now properly handle shutdown of log forwarders, so an ObserveService client does not prevent auraed from exiting and shuts down properly.

[dmah42]

part of a different change?

[mccormickt]

Happy to split it out if you'd like, but this was another fix that I encountered and is required for this change to work as well. I missed this in #601, and not clearing readiness caused the Tokio worker pool to be consumed by log reader tasks. This would block any gRPC startup tasks from making progress and caused a livelock.