[LPT] Support FakeQuantize with convert on intervals (#9579)

* [LPT] Support FakeQuantize with convert on intervals

* [LPT] GPU tests

src/common/low_precision_transformations/include/low_precision/quantization_details.hpp

@@ -29,7 +29,7 @@ public:
             const std::vector<float>& outputLowValues,
             const std::vector<float>& outputHighValues);
 
-    static bool outputLayoutIsSupported(std::shared_ptr<opset1::FakeQuantize> quantize);
+    static bool outputLayoutIsSupported(std::shared_ptr<opset1::FakeQuantize> quantize, bool isConvertExpected = false);
 
     static void getInputIntervals(
             std::shared_ptr<opset1::FakeQuantize> quantize,

src/common/low_precision_transformations/src/low_precision.cpp

@@ -275,7 +275,7 @@ bool ngraph::pass::low_precision::LowPrecision::isFunctionQuantized(const std::s
 
             const std::shared_ptr<ngraph::opset1::FakeQuantize> fakeQuantize = ov::as_type_ptr<ngraph::opset1::FakeQuantize>(parent);
             if ((fakeQuantize != nullptr) &&
-                QuantizationDetails::outputLayoutIsSupported(fakeQuantize) &&
+                QuantizationDetails::outputLayoutIsSupported(fakeQuantize, true) &&
                 QuantizationDetails::isSupportedLevel(fakeQuantize->get_levels())) {
                 return true;
             }

src/common/low_precision_transformations/src/quantization_details.cpp

@@ -49,11 +49,19 @@ QuantizationDetails::QuantizationDetails(const size_t levels, const std::vector<
       outputLowValues(outputLowValues),
       outputHighValues(outputHighValues) {}
 
-bool QuantizationDetails::outputLayoutIsSupported(std::shared_ptr<opset1::FakeQuantize> quantize) {
-    return ov::is_type<opset1::Constant>(quantize->get_input_node_ptr(1)) &&
-        ov::is_type<opset1::Constant>(quantize->get_input_node_ptr(2)) &&
-        ov::is_type<opset1::Constant>(quantize->get_input_node_ptr(3)) &&
-        ov::is_type<opset1::Constant>(quantize->get_input_node_ptr(4));
+bool QuantizationDetails::outputLayoutIsSupported(std::shared_ptr<opset1::FakeQuantize> quantize, bool isConvertExpected) {
+    const auto inputs = quantize->inputs();
+    for (size_t i = 1; i < inputs.size(); ++i) {
+        const auto node = inputs[i].get_source_output().get_node_shared_ptr();
+        bool supported = ov::is_type<opset1::Constant>(node);
+        if (!supported && isConvertExpected) {
+            supported = ov::is_type<op::Convert>(node) && ov::is_type<opset1::Constant>(node->get_input_node_ptr(0));
+        }
+        if (!supported) {
+            return false;
+        }
+    }
+    return true;
 }
 
 void QuantizationDetails::getInputIntervals(