sample: improve ollama engine sampler performance (#9374)

This change bring in various interface cleanups along with greatly improving the performance of the sampler.

Tested with llama3.2 on local machine.
Improves performance from ~ 70 tokens/s -> 135 tokens/s with topK(40) enabled.
Without topK performance is ~ 110 tokens/s

runner/ollamarunner/runner.go

@@ -589,11 +589,19 @@ func (s *Server) completion(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 
+	sampler := sample.NewSampler(
+		req.Temperature,
+		req.TopK,
+		req.TopP,
+		req.MinP,
+		req.Seed,
+	)
+
 	seq, err := s.NewSequence(req.Prompt, req.Images, NewSequenceParams{
 		numPredict: req.NumPredict,
 		stop:       req.Stop,
 		numKeep:    int32(req.NumKeep),
-		sampler:    sample.Greedy(), // TODO: add support for different samplers when performance is optimized
+		sampler:    sampler,
 		embedding:  false,
 	})
 	if err != nil {