[ruby] improve handling for recv_message op failures (#34016)

[ruby] improve handling for recv_message op failures (don't assume recv_message ops won't fail e.g. if the call was cancelled by the peer)

src/ruby/lib/grpc/generic/active_call.rb

@@ -261,6 +261,9 @@ module GRPC
         @metadata_received = true
       end
       get_message_from_batch_result(batch_result)
+    rescue GRPC::Core::CallError => e
+      GRPC.logger.info("remote_read: #{e}")
+      nil
     end
 
     def get_message_from_batch_result(recv_message_batch_result)
@@ -327,14 +330,7 @@ module GRPC
     def each_remote_read_then_finish
       return enum_for(:each_remote_read_then_finish) unless block_given?
       loop do
-        resp =
+        resp = remote_read
-          begin
-            remote_read
-          rescue GRPC::Core::CallError => e
-            GRPC.logger.warn("In each_remote_read_then_finish: #{e}")
-            nil
-          end
-
         break if resp.nil?  # the last response was received
         yield resp
       end

src/ruby/spec/generic/rpc_server_spec.rb

@@ -668,9 +668,9 @@ describe GRPC::RpcServer do
       def check_multi_req_view_of_finished_call(call)
         common_check_of_finished_server_call(call)
 
-        expect do
+        l = []
-          call.each_remote_read.each { |r| p r }
+        call.each_remote_read.each { |r| l << r }
-        end.to raise_error(GRPC::Core::CallError)
+        expect(l.size).to eq(0)
       end
 
       def common_check_of_finished_server_call(call)