Detect inline cycles faster

E.g. in the following code

    fun x() {}
    inline fun f() { x(); g() }
    inline fun g() { x(); f() }

the old implementation of inline cycle detection bailed out after
generating 3 calls of x() in each function, while the new one stops
after 2. In other words, code generation for a single function is no
longer reentered.

compiler/testData/diagnostics/testsWithJvmBackend/suspendInlineCycle.kt

@@ -0,0 +1,11 @@
+// !RENDER_DIAGNOSTICS_FULL_TEXT
+// Note: 4 diagnostics per call because there are 2 synthetic $$forInline methods.
+suspend inline fun inlineFun1(p: () -> Unit) {
+    p()
+    <!INLINE_CALL_CYCLE, INLINE_CALL_CYCLE, INLINE_CALL_CYCLE, INLINE_CALL_CYCLE!>inlineFun2(p)<!>
+}
+
+suspend inline fun inlineFun2(p: () -> Unit) {
+    p()
+    <!INLINE_CALL_CYCLE, INLINE_CALL_CYCLE, INLINE_CALL_CYCLE, INLINE_CALL_CYCLE!>inlineFun1(p)<!>
+}