feat: wire persist_init/recover/start/stop into cluster lifecycle (P0-F)

- Add `persist_t persist` field to cluster_t (cluster_def.h includes persist.h)
- Add `workdir` and `name` fields to myconf_t; main.c passes them through
- cluster_init: call persist_set_workdir, persist_init, persist_recover after apiserver_init
- cluster_start: call persist_start (launches fsync thread + registers WAL hooks)
- cluster_fini: call persist_stop + persist_fini before delaymgr/vtaskmgr teardown
- persist_init: store cluster back-pointer (replaces (void)c stub)
- persist_start: register persist_job/queue/worker_hook_fun via xce_add_hook when cluster != NULL
- Add test_cluster_persist.c: compile-time struct check + config defaults + workdir path derivation
- Register test in controller_test.c and meson.build

All controller tests pass (1/1 OK, 50s).

src/include/xce/cluster_def.h

@@ -14,6 +14,7 @@
 #include "myconf.h"
 #include "mylock.h"
 #include "netio.h"
+#include "persist.h"
 #include "vtaskmgr.h"
 #include "xce/types.h"
 #include "xlist.h"
@@ -114,6 +115,7 @@ typedef struct cluster_s {
     vtaskmgr_t          vtaskmgr;
     cbuf_t              wbuf;
     int                 running;
+    persist_t           persist;   /* Persistence module — P0-F */
 } cluster_t;
 
 #endif /* XCE_CLUSTER_DEF_H */