refactor(server): plumb Arc<Settings> through workers and routes

Config sheds its 14 Bucket-B fields (retention, whisper, ollama, llm) and
gains a sibling Arc<Settings> in AppState, loaded from settings.toml at
startup via Settings::load_or_default. Workers (transcribe, analyze,
recovery, auto_trigger) and routes (health, bulk, case_actions,
user_web) take settings as a separate parameter or State<> extractor;
Config::llm_configured moves to Settings::llm_configured.

TestConfig::build_pair() returns (Arc<Config>, Arc<Settings>); the new
create_router_with_settings / create_router_and_session_store_with_settings
helpers wire both into integration tests that exercise LLM/Whisper/Ollama.
The legacy single-Arc create_router falls back to Settings::default()
so the 17 non-LLM tests stay untouched.

Verified: cargo build clean, clippy --all-targets clean, 318 tests pass.

server/src/transcribe/recovery.rs

@@ -4,11 +4,11 @@ use doctate_common::oneliners::OnelinerState;
 use tracing::{info, warn};
 
 use super::{TranscribeJob, TranscribeSender};
-use crate::config::Config;
 use crate::events::EventSender;
 use crate::gazetteer::Gazetteer;
 use crate::oneliner_locks::OnelinerLocks;
 use crate::paths;
+use crate::settings::Settings;
 
 /// Walk `data_path/*/` and enqueue every recording pending transcription for
 /// every user. Intended for startup; the same primitive backs the per-user
@@ -204,7 +204,7 @@ async fn has_any_transcript(case_dir: &Path) -> bool {
 pub async fn regenerate_missing_oneliners(
     data_path: &Path,
     client: &reqwest::Client,
-    config: &Config,
+    settings: &Settings,
     vocab: &Gazetteer,
     events_tx: &EventSender,
     locks: &OnelinerLocks,
@@ -215,7 +215,7 @@ pub async fn regenerate_missing_oneliners(
     }
     info!(count = cases.len(), "Regenerating missing oneliners");
     for (case_dir, slug) in cases {
-        super::worker::update_oneliner(&case_dir, &slug, client, config, vocab, events_tx, locks)
+        super::worker::update_oneliner(&case_dir, &slug, client, settings, vocab, events_tx, locks)
             .await;
     }
 }
@@ -228,7 +228,7 @@ pub async fn regenerate_missing_oneliners_for_user(
     user_root: &Path,
     slug: &str,
     client: &reqwest::Client,
-    config: &Config,
+    settings: &Settings,
     vocab: &Gazetteer,
     events_tx: &EventSender,
     locks: &OnelinerLocks,
@@ -238,7 +238,7 @@ pub async fn regenerate_missing_oneliners_for_user(
         return;
     }
     for case_dir in cases {
-        super::worker::update_oneliner(&case_dir, slug, client, config, vocab, events_tx, locks)
+        super::worker::update_oneliner(&case_dir, slug, client, settings, vocab, events_tx, locks)
             .await;
     }
 }

server/src/transcribe/worker.rs

@@ -14,15 +14,18 @@ use crate::events::{self, CaseEventKind, EventSender};
 use crate::gazetteer::Gazetteer;
 use crate::oneliner_locks::OnelinerLocks;
 use crate::paths;
+use crate::settings::Settings;
 use crate::{BusyGuard, WorkerBusy};
 
 const ONELINER_TIMEOUT: Duration = Duration::from_secs(60);
 
 /// Consume transcription jobs sequentially. Whisper holds the GPU, so parallelism
 /// here would only cause contention. One job in flight at a time.
+#[allow(clippy::too_many_arguments)]
 pub async fn run(
     mut rx: TranscribeReceiver,
     config: Arc<Config>,
+    settings: Arc<Settings>,
     client: reqwest::Client,
     worker_busy: WorkerBusy,
     vocab: Arc<Gazetteer>,
@@ -30,7 +33,7 @@ pub async fn run(
     oneliner_locks: OnelinerLocks,
 ) {
     info!("Transcription worker started");
-    let timeout = Duration::from_secs(config.whisper_timeout_seconds);
+    let timeout = Duration::from_secs(settings.whisper.timeout_seconds);
 
     while let Some(job) = rx.recv().await {
         let _guard = BusyGuard::new(worker_busy.clone());
@@ -44,7 +47,7 @@ pub async fn run(
         // Look up per-user Whisper settings live from the shared config so that
         // edits to users.toml (on next restart) reach in-flight jobs naturally.
         // Unknown slug → empty settings (service falls back to language=de).
-        let settings: WhisperUserSettings = config
+        let user_whisper: WhisperUserSettings = config
             .users
             .iter()
             .find(|u| u.slug == job.user_slug)
@@ -70,10 +73,10 @@ pub async fn run(
 
         let text = match whisper::transcribe(
             &client,
-            &config.whisper_url,
+            &settings.whisper.url,
             remuxed.path(),
             timeout,
-            &settings,
+            &user_whisper,
         )
         .await
         {
@@ -139,7 +142,7 @@ pub async fn run(
                 case_dir,
                 &job.user_slug,
                 &client,
-                &config,
+                &settings,
                 &vocab,
                 &events_tx,
                 &oneliner_locks,
@@ -220,7 +223,7 @@ pub async fn update_oneliner(
     case_dir: &Path,
     user_slug: &str,
     client: &reqwest::Client,
-    config: &Config,
+    settings: &Settings,
     vocab: &Gazetteer,
     events_tx: &EventSender,
     locks: &OnelinerLocks,
@@ -302,9 +305,9 @@ pub async fn update_oneliner(
     // during the call.
     let state = match ollama::generate_oneliner(
         client,
-        &config.ollama_url,
-        &config.ollama_model,
-        config.ollama_keep_alive,
+        &settings.ollama.url,
+        &settings.ollama.model,
+        settings.ollama.keep_alive,
         &transcript,
         ONELINER_TIMEOUT,
     )