fix(ruleset): avoid async fetch limiter deadlock

base/pref.example.ini

@@ -308,4 +308,4 @@ coalesce_retry_on_5xx=true
 ;0 disables completed response caching. Values above 5 seconds are clamped to 5.
 response_cache_ttl=0
 ;0 disables the limiter. A small value is safer on single-core deployments.
-max_async_fetches=2
+max_async_fetches=0

base/pref.example.toml

@@ -366,4 +366,4 @@ coalesce_retry_on_5xx = true
 # 0 disables completed response caching. If enabled, values above 5 seconds are clamped to 5.
 response_cache_ttl = 0
 # 0 disables the limiter. A small value is safer on single-core deployments.
-max_async_fetches = 2
+max_async_fetches = 0

base/pref.example.yml

@@ -175,4 +175,4 @@ advanced:
   enable_request_coalescing: true
   coalesce_retry_on_5xx: true
   response_cache_ttl: 0
-  max_async_fetches: 2
+  max_async_fetches: 0

src/handler/multithread.cpp

@@ -129,8 +129,10 @@ std::shared_future<std::string> fetchFileAsync(const std::string &path, const st
         retVal = std::async(std::launch::async, [path](){return fileGet(path, true);});
     else if(isLink(path))
     {
-        auto slot = acquireAsyncFetchSlot();
-        retVal = std::async(std::launch::async, [path, proxy, cache_ttl, context, slot](){return webGet(path, proxy, cache_ttl, nullptr, nullptr, context);});
+        retVal = std::async(std::launch::async, [path, proxy, cache_ttl, context](){
+            auto slot = acquireAsyncFetchSlot();
+            return webGet(path, proxy, cache_ttl, nullptr, nullptr, context);
+        });
     }
     else
         return make_ready_future(std::string());

src/handler/settings.h

@@ -79,7 +79,7 @@ struct Settings {
 
   // request coalescing and bounded background work
   bool enableRequestCoalescing = true, coalesceRetryOn5xx = true;
-  int responseCacheTtl = 0, maxAsyncFetches = 2;
+  int responseCacheTtl = 0, maxAsyncFetches = 0;
   unsigned long long configGeneration = 0;
 
   // limits