Replace go-openai with native client

Drop the github.com/sashabaranov/go-openai dependency in favor of an
internal HTTP client.

The LLM package is now split into:
- llm.go: type definitions (Message, ChatRequest, ChatResponse, etc.)
- client.go: client construction and configuration
- client_completions.go: Completions and CompletionsStream methods
- client_models.go: ListModels
- client_tools.go: tool call execution

CompletionsStream uses typed SSE events (delta, done, message) so the
frontend can render progressive streaming output.

Move tool.go and tool_test.go from internal/tool/ into internal/llm/
to co-locate the registry with the client that consumes it.

Update job.go and service.go to use the new llm.Message types and
Completions/CompletionsStream methods. Remove the Voice field from
chat request/response (voice is now tracked via message metadata).

Frontend: handle delta/done/message SSE events, render a streaming
placeholder that receives progressive text deltas, and finalize with
full message binding (reasoning, tool calls, actions).

internal/llm/client_test.go

@@ -0,0 +1,132 @@
+package llm
+
+import (
+	"testing"
+)
+
+func TestConfigValidate(t *testing.T) {
+	tests := []struct {
+		name    string
+		cfg     Config
+		wantErr bool
+	}{
+		{
+			name:    "empty config",
+			cfg:     Config{},
+			wantErr: true,
+		},
+		{
+			name:    "missing URL",
+			cfg:     Config{},
+			wantErr: true,
+		},
+		{
+			name: "invalid timeout",
+			cfg: Config{
+				URL:     "http://localhost:8080",
+				Timeout: "not-a-duration",
+			},
+			wantErr: true,
+		},
+		{
+			name: "valid minimal config",
+			cfg: Config{
+				URL: "http://localhost:8080",
+			},
+			wantErr: false,
+		},
+		{
+			name: "valid config with timeout",
+			cfg: Config{
+				URL:     "http://localhost:8080",
+				Timeout: "15m",
+			},
+			wantErr: false,
+		},
+		{
+			name: "valid config with complex timeout",
+			cfg: Config{
+				URL:     "http://localhost:8080",
+				Timeout: "1h30m45s",
+			},
+			wantErr: false,
+		},
+		{
+			name: "valid full config",
+			cfg: Config{
+				Key:           "sk-test-key",
+				SystemMessage: "You are a helpful assistant.",
+				Timeout:       "30m",
+				URL:           "http://localhost:8080",
+			},
+			wantErr: false,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			err := tt.cfg.Validate()
+			if (err != nil) != tt.wantErr {
+				t.Errorf(
+					"Validate() error = %v, wantErr %v",
+					err, tt.wantErr,
+				)
+			}
+		})
+	}
+}
+
+func TestNewClient(t *testing.T) {
+	tests := []struct {
+		name    string
+		cfg     Config
+		wantErr bool
+	}{
+		{
+			name:    "invalid config",
+			cfg:     Config{},
+			wantErr: true,
+		},
+		{
+			name: "valid config without timeout",
+			cfg: Config{
+				URL: "http://localhost:8080",
+			},
+			wantErr: false,
+		},
+		{
+			name: "valid config with timeout",
+			cfg: Config{
+				URL:     "http://localhost:8080",
+				Timeout: "10m",
+			},
+			wantErr: false,
+		},
+		{
+			name: "valid config with all fields",
+			cfg: Config{
+				Key:           "test-key",
+				SystemMessage: "Test message",
+				Timeout:       "5m",
+				URL:           "http://localhost:8080",
+			},
+			wantErr: false,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			client, err := NewClient(tt.cfg, nil, nil)
+			if (err != nil) != tt.wantErr {
+				t.Errorf(
+					"NewClient() error = %v, wantErr %v",
+					err, tt.wantErr,
+				)
+				return
+			}
+			if !tt.wantErr && client == nil {
+				t.Error("NewClient() returned nil client")
+			}
+		})
+	}
+}