never add rts to inline asmsubs and always inline them regardless of optimization setting

otherwise they can't specify a sequence of assembly instructions that should be inserted in-place, such as those that manipulate the cpu stack.
for instance cx16.irqsafe_set_irqd() / cx16.irqsafe_clear_irqd()

codeGenCpu6502/src/prog8/codegen/cpu6502/ProgramAndVarsGen.kt

@@ -271,9 +271,7 @@ internal class ProgramAndVarsGen(
 
     internal fun translateAsmSubroutine(sub: PtAsmSub) {
         if(sub.inline) {
-            if(options.optimize) {
-                return
-            }
+            return      // subroutine gets inlined at call site.
         }
 
         asmgen.out("")