Plumb ctx through classifier and Ollama HTTP path

OllamaClient.Generate/GenerateText/call now take ctx and build the HTTP
request via NewRequestWithContext, so an in-flight LLM call is aborted
when the parent context is cancelled (Ctrl-C). Classifier.Classify and
its tier helpers take ctx too. ProcessFunc gets a ctx parameter so the
poller can forward its cancellable context down to classification.

Explainer.summarize manages its own 60s context since reaction-driven
flow has no parent ctx to inherit.

internal/ingestion/poller.go

@@ -12,7 +12,9 @@ import (
 )
 
 // ProcessFunc is called for each new feed item that needs classification.
-type ProcessFunc func(item *FeedItem)
+// ctx is the poller's parent context; callbacks should pass it into any
+// downstream LLM/HTTP work so shutdown aborts in-flight calls.
+type ProcessFunc func(ctx context.Context, item *FeedItem)
 
 // Poller manages per-source RSS polling goroutines.
 type Poller struct {
@@ -182,7 +184,7 @@ func (p *Poller) pollOnceWithErr(ctx context.Context, src config.SourceConfig) e
 			continue
 		}
 
-		p.process(&items[i])
+		p.process(ctx, &items[i])
 		newCount++
 	}
 
@@ -212,7 +214,7 @@ func (p *Poller) pollOnceWithErr(ctx context.Context, src config.SourceConfig) e
 			continue
 		}
 
-		p.process(&FeedItem{
+		p.process(ctx, &FeedItem{
 			GUID:        s.GUID,
 			Headline:    s.Headline,
 			Lede:        s.Lede,