Clean up runtime cancellation and logging follow-ups

Address pre-existing follow-ups surfaced during PR review:

- TransmissionPolicyManager: fix three LOG_TRACE calls that used %d for
  uint64_t / chrono::milliseconds::rep values. The negative-delay log was
  already corrected; this extends the same fix to the WAIT / SCHED /
  reschedule traces so all delay logging is portable.

- WorkerThread: remove the dead 'count' member. It was incremented in
  Join() and Queue() but never read or decremented anywhere.

- HttpClient_Apple::CancelAllRequests: cancel each in-flight request
  under a single critical section instead of collecting ids, releasing
  the mutex, and calling CancelRequestAsync per id. The previous pattern
  opened a small window where the request could complete and be erased
  between the two locked sections. The spin-wait that follows now also
  meaningfully blocks until each NSURLSession completion handler has
  fired and the corresponding HttpRequestApple destructor has removed
  itself via HttpClient_Apple::Erase, which is the actual signal that
  cancellation has settled.

Validation:
- Host UnitTests on macOS arm64: 488/488 pass.
- TransmissionPolicyManagerTests + HttpClientManagerTests +
  HttpResponseDecoderTests --gtest_repeat=10: 46/46 each round.
- iphonesimulator and iphoneos -fsyntax-only of HttpClient_Apple.mm
  both clean.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Author: bmehta001

lib/http/HttpClient_Apple.mm

@@ -186,17 +186,19 @@ void Cancel()
 
 void HttpClient_Apple::CancelAllRequests()
 {
-    std::vector<std::string> ids;
     {
         std::lock_guard<std::mutex> lock(m_requestsMtx);
-        for (auto const& item : m_requests) {
-            ids.push_back(item.first);
+        for (auto const& item : m_requests)
+        {
+            auto* request = static_cast<HttpRequestApple*>(item.second);
+            if (request != nullptr)
+            {
+                LOG_TRACE("HTTP request=%p id=%s being aborted...", request, item.first.c_str());
+                request->Cancel();
+            }
         }
     }
 
-    for (const auto &id : ids)
-        CancelRequestAsync(id);
-
     while (true)
     {
         {

lib/pal/WorkerThread.cpp

@@ -38,7 +38,6 @@ namespace PAL_NS_BEGIN {
         Event                 m_event;
         MAT::Task*            m_itemInProgress;
         bool                  m_shuttingDown = false;
-        int count = 0;
 
     public:
 
@@ -63,7 +62,6 @@ namespace PAL_NS_BEGIN {
                 if (!m_shuttingDown) {
                     m_shuttingDown = true;
                     m_queue.push_back(new WorkerThreadShutdownItem());
-                    count++;
                     m_event.post();
                 }
             }
@@ -126,7 +124,6 @@ namespace PAL_NS_BEGIN {
             else {
                 m_queue.push_back(item);
             }
-            count++;
             m_event.post();
         }
 

lib/tpm/TransmissionPolicyManager.cpp

@@ -164,7 +164,9 @@ namespace MAT_NS_BEGIN {
                 // Don't need to cancel and reschedule if it's about to happen now anyways.
                 // the completion of upload will schedule more uploads as-needed, we only
                 // want to avoid the unnecessary wasteful rescheduling.
-                LOG_TRACE("WAIT  upload %d ms for lat=%d", delta, m_runningLatency);
+                LOG_TRACE("WAIT  upload %llu ms for lat=%d",
+                    static_cast<unsigned long long>(delta),
+                    static_cast<int>(m_runningLatency));
                 return;
             }
         }
@@ -200,7 +202,9 @@ namespace MAT_NS_BEGIN {
             m_isUploadScheduled = true;
             m_scheduledUploadTime = PAL::getMonotonicTimeMs() + delay.count();
             m_runningLatency = latency;
-            LOG_TRACE("SCHED upload %d ms for lat=%d", delay.count(), m_runningLatency);
+            LOG_TRACE("SCHED upload %lld ms for lat=%d",
+                static_cast<long long>(delay.count()),
+                static_cast<int>(m_runningLatency));
             m_scheduledUpload = PAL::scheduleTask(&m_taskDispatcher, static_cast<unsigned>(delay.count()), this, &TransmissionPolicyManager::uploadAsync, latency);
         }
     }
@@ -264,7 +268,8 @@ namespace MAT_NS_BEGIN {
         // Rescheduling upload
         if (nextUpload.count() >= 0)
         {
-            LOG_TRACE("Scheduling upload in %d ms", nextUpload.count());
+            LOG_TRACE("Scheduling upload in %lld ms",
+                static_cast<long long>(nextUpload.count()));
             EventLatency proposed = calculateNewPriority();
             scheduleUpload(nextUpload, proposed); // reschedule uploadAsync again
         }