Update boost::asio usage to conform to newer standards: (#144)

* Convert boost::asio::io_service to boost::asio::io_context
  * Convert strand.wrap(...) to boost::asio::bind_executor(strand, ...)
  * Convert strand.dispatch(...) to boost::asio::dispatch(strand, ...)
  * Convert io_context.reset() to io_context.restart()
  * Convert null_buffers() usage to socket.async_wait(...)
  * Drop usage of GET_IO_SERVICE macro from monero
  * Refactor REST server to manage resources better

src/scanner.cpp

@@ -1080,7 +1080,7 @@ namespace lws
       {
         // Run possible SIGINT handler
         self.io_.poll_one();
-        self.io_.reset();
+        self.io_.restart();
         if (self.has_shutdown())
           return {lws::error::signal_abort_process};
 
@@ -1098,7 +1098,7 @@ namespace lws
       {
         // Run possible SIGINT handler
         self.io_.poll_one();
-        self.io_.reset();
+        self.io_.restart();
         if (self.has_shutdown())
           return {lws::error::signal_abort_process};
 
@@ -1334,11 +1334,11 @@ namespace lws
       thread_count = std::max(std::size_t(1), thread_count);
 
     /*! \NOTE Be careful about references and lifetimes of the callbacks. The
-      ones below are safe because no `io_service::run()` call is after the
+      ones below are safe because no `io_context::run()` call is after the
       destruction of the references.
 
       \NOTE That `ctx` will need a strand or lock if multiple
-        `io_service::run()` calls are used. */
+        `io_context::run()` calls are used. */
 
     boost::asio::steady_timer rate_timer{sync_.io_};
     class rate_updater