feat(scheduler-core): implement builtin_fcfs session and post-allocation hooks (task 2.3)

Adds begin_session/next_job/end_session/on_dispatch/on_pend, driving a
dual-cursor walk (queue then job) over builtin_fcfs's private ordq_t
state. next_job() returns candidates in (queue priority desc, job
priority desc, queue_time asc) order and NULL once exhausted;
on_dispatch removes permanently, on_pend is a true no-op.

Review flagged an unsynchronized cross-thread access risk between
next_job() (picker thread) and on_dispatch/on_pend (allocator thread)
once wired live in task 2.4 — recorded in tasks.md Implementation Notes
for that task to close before registering builtin_fcfs as the default.

Requirements: 3.2, 3.4, 3.6

src/scheduler/testsuite/test_picker_fcfs.c

@@ -32,6 +32,27 @@
 #include "plugin.h"
 #include "sched/types.h"
 
+/* builtin_fcfs.c's next_job() (task 2.3) calls sched_queue_is_ready()/
+ * sched_job_is_ready() (declared in scheduler.h, defined in scheduler.c),
+ * which this translation unit does not link -- same as every other
+ * white-box picker test in this directory that #includes builtin_fcfs.c
+ * directly. This file's own tests (tasks 2.1/2.2) never call next_job(), so
+ * a simple always-ready stub (mirroring scheduler.c's own current stub,
+ * scheduler.c:1267/1270) is enough to satisfy the linker;
+ * test_picker_fcfs_session.c is where the real skip-when-not-ready behavior
+ * is exercised, via controllable test doubles. */
+int sched_queue_is_ready(xce_sched_queue_t* q)
+{
+    (void)q;
+    return 1;
+}
+
+int sched_job_is_ready(xce_sched_job_t* j)
+{
+    (void)j;
+    return 1;
+}
+
 #include "../src/plugins/builtin_fcfs.c"
 
 void setUp(void) { return; }