JVM IR: clear BindingContext after psi2ir

This helps to reduce peak memory in lowerings/codegen by a lot.

A more robust approach would be to have a separate BindingContext for
each file, and clear each of them after running psi2ir on it. This would
also lower peak memory usage in psi2ir.

Provide a fallback workaround compiler argument
-Xir-do-not-clear-binding-context just in case BindingContext is in fact
used somewhere and it's not caught by tests.

compiler/tests-common/tests/org/jetbrains/kotlin/checkers/BaseDiagnosticsTest.kt

@@ -34,10 +34,7 @@ import org.jetbrains.kotlin.checkers.diagnostics.factories.SyntaxErrorDiagnostic
 import org.jetbrains.kotlin.checkers.utils.CheckerTestUtil
 import org.jetbrains.kotlin.checkers.utils.DiagnosticsRenderingConfiguration
 import org.jetbrains.kotlin.cli.jvm.compiler.KotlinCoreEnvironment
-import org.jetbrains.kotlin.config.JvmTarget
-import org.jetbrains.kotlin.config.LanguageFeature
-import org.jetbrains.kotlin.config.LanguageVersionSettings
-import org.jetbrains.kotlin.config.LanguageVersionSettingsImpl
+import org.jetbrains.kotlin.config.*
 import org.jetbrains.kotlin.descriptors.DeclarationDescriptor
 import org.jetbrains.kotlin.descriptors.impl.ModuleDescriptorImpl
 import org.jetbrains.kotlin.diagnostics.*
@@ -269,6 +266,8 @@ abstract class BaseDiagnosticsTest : KotlinMultiFileTestWithJava<TestModule, Tes
                     platform = null,
                     withNewInference,
                     languageVersionSettings,
+                    // When using JVM IR, binding context is empty at the end of compilation, so debug info markers can't be computed.
+                    environment.configuration.getBoolean(JVMConfigurationKeys.IR),
                 ),
                 DataFlowValueFactoryImpl(languageVersionSettings),
                 moduleDescriptor,