compact: retain newest ~20% of context; refactor startup into internal/app

- Replace fixed 6-message compact retention with ratio-based logic:
  retain ~20% of compactable estimated tokens, with a 6-message floor
  and user-turn boundary alignment
- Add compactRetainedTokenBudget, compactRecentStart, compactTurnBoundary
  helpers in internal/agent/agent.go
- Add TestAgentCompactKeepsTwentyPercentRecentContext
- Move startup/session/repl/bootstrap logic from root into internal/app
  so main.go is a thin binary entry point (15 lines)
- Update /compact docs in README.md

internal/app/repl.go

@@ -0,0 +1,60 @@
+package app
+
+import (
+	"bufio"
+	"context"
+	"fmt"
+	"os"
+	"strings"
+
+	"agentu/internal/agent"
+	"agentu/internal/session"
+)
+
+func repl(ctx context.Context, assistant *agent.Agent, modelManager *session.Manager) error {
+	scanner := bufio.NewScanner(os.Stdin)
+	scanner.Buffer(make([]byte, 0, 64*1024), 1024*1024)
+
+	fmt.Println("agentu")
+	fmt.Println("Type /exit to quit, /clear to reset context, /compact to compress context.")
+	for {
+		fmt.Print("> ")
+		if !scanner.Scan() {
+			if err := scanner.Err(); err != nil {
+				return err
+			}
+			fmt.Println()
+			printExitSession(modelManager)
+			return nil
+		}
+
+		input := strings.TrimSpace(scanner.Text())
+		switch input {
+		case "":
+			continue
+		case "/exit", "/quit":
+			printExitSession(modelManager)
+			return nil
+		case "/clear":
+			assistant.Clear()
+			fmt.Println("context cleared")
+			continue
+		case "/compact":
+			if err := assistant.Compact(ctx); err != nil {
+				fmt.Fprintln(os.Stderr, "compact error:", err)
+			} else {
+				fmt.Println("context compacted")
+				if modelManager != nil {
+					_ = modelManager.Save()
+				}
+			}
+			continue
+		}
+
+		if err := assistant.RunTurn(ctx, input, os.Stdout, os.Stderr); err != nil {
+			fmt.Fprintln(os.Stderr, "\nerror:", err)
+		}
+		_ = modelManager.Save()
+		fmt.Println()
+	}
+}