ipn/ipnlocal, tka: compact TKA state after every sync

Previously a TKA compaction would only run when a node starts, which means a long-running node could use unbounded storage as it accumulates ever-increasing amounts of TKA state. This patch changes TKA so it runs a compaction after every sync.

Updates https://github.com/tailscale/corp/issues/33537

Change-Id: I91df887ea0c5a5b00cb6caced85aeffa2a4b24ee
Signed-off-by: Alex Chan <alexc@tailscale.com>

tstest/chonktest/tailchonk_test.go

@@ -18,7 +18,7 @@ func TestImplementsChonk(t *testing.T) {
 		{
 			name: "Mem",
 			newChonk: func(t *testing.T) tka.Chonk {
-				return &tka.Mem{}
+				return tka.ChonkMem()
 			},
 		},
 		{
@@ -42,7 +42,7 @@ func TestImplementsCompactableChonk(t *testing.T) {
 		{
 			name: "Mem",
 			newChonk: func(t *testing.T) tka.CompactableChonk {
-				return &tka.Mem{}
+				return tka.ChonkMem()
 			},
 		},
 		{