fix(libxce-failover): classify and propagate netio I/O errors on failover

Add io_classify_io_error and propagate endpoint/epoll errors through
endpoint_update_epoll and cbuf_get instead of swallowing them, so client
failover surfaces the real error rather than hanging. Adds netio arm-failure,
close-propagation, and request send-propagation tests.

Pre-existing uncommitted work, committed to establish a clean baseline before
ha-commit-durability implementation.

src/lib/testsuite/test_request_send_propagation.c

@@ -0,0 +1,72 @@
+/*
+ * test_request_send_propagation.c — bug-650: xce_request() must PROPAGATE an
+ * endpoint_enq() send failure, not swallow it.
+ *
+ * xce_request() (and the other request.c senders) ignored endpoint_enq()'s
+ * return and always returned 0.  So when the transport was dead — endpoint_enq()
+ * returns <0 and marks the endpoint dead (bug-646) — xce_request() still
+ * reported success.  xce_jobsubmit() then fell through to a BLOCKING cbuf_get()
+ * and hung forever waiting for a reply to a request that was never transmitted
+ * (server idle, client fd unconnected).  xce_request() must return the negative
+ * enq error so the failover loop rotates instead of blocking.
+ */
+
+#define _GNU_SOURCE
+#include <errno.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/socket.h>
+#include <unistd.h>
+
+#include <unity/unity.h>
+#include "netio.h"
+#include "request.h"
+
+/*
+ * A request that ENCODES successfully but whose send fails because the endpoint
+ * is dead (EPOLL_CTL_MOD on an fd that is not registered → ENOENT) must make
+ * xce_request() return a negative error, not 0.
+ */
+static void test_xce_request_propagates_send_failure(void)
+{
+    iomgr_t iomgr;
+    memset(&iomgr, 0, sizeof(iomgr));
+    TEST_ASSERT_EQUAL_INT(0, iomgr_init(&iomgr));
+
+    endpoint_t *ep = endpoint_get();
+    TEST_ASSERT_NOT_NULL(ep);
+
+    ep->fd = socket(AF_INET, SOCK_STREAM, 0); /* valid fd, NOT in the epoll */
+    TEST_ASSERT_TRUE(ep->fd >= 0);
+    ep->iomgr              = &iomgr;
+    ep->ssl_handshake_done = 1;
+    ep->flags              = 0; /* force iomgr_update_poll MOD on the unregistered fd */
+
+    char           nm[] = "tester";
+    char           ky[] = "key";
+    xce_peerinfo_t peer = { .name = nm, .key = ky };
+    ep->peer            = &peer;
+
+    xce_request_t req;
+    memset(&req, 0, sizeof(req));
+    req.opcode = OPCode_nop;
+    req.seq    = 42;
+
+    int rc = xce_request(ep, &req);
+
+    /* Old code: always 0 (swallowed). Fixed: the negative enq error propagates. */
+    TEST_ASSERT_TRUE(rc < 0);
+
+    endpoint_fini(ep);
+    free(ep);
+    close(iomgr.epoll_fd);
+    printf("test_xce_request_propagates_send_failure: PASS\n");
+}
+
+void test_request_send_propagation(void)
+{
+    printf("\n>> request send-failure propagation tests\n\n");
+
+    RUN_TEST(test_xce_request_propagates_send_failure);
+}