Make malign-double effective only for x86

[MaskRay]

Summary

Making -malign-double effective only for x86.

Fixes.
https://bugs.llvm.org/show_bug.cgi?id=45136

[MaskRay]

-Xclang -malign-double passes a CC1 option. This is like poking into the compiler internals. The interface is not stable, and some flags may not make sense (they can be precluded by clang/lib/Driver logic), so I am not sure why it is reported as a bug in the first place.

edit: The problem is that Opts.AlignDouble in TargetInfo.cpp needs some adaptation to make it work for aarch64_32. I will on a patch.

I think we should just add aarch64_32 support when it becomes clearer how aarch64_32 needs -malign-double.

clang/lib/Frontend/CompilerInvocation.cpp

Opts.ObjCConstantStringClass =
       std::string(Args.getLastArgValue(OPT_fconstant_string_class));
   Opts.ObjCDefaultSynthProperties =
     !Args.hasArg(OPT_disable_objc_default_synthesize_properties);
   Opts.EncodeExtendedBlockSig =
     Args.hasArg(OPT_fencode_extended_block_signature);
   Opts.EmitAllDecls = Args.hasArg(OPT_femit_all_decls);
   Opts.PackStruct = getLastArgIntValue(Args, OPT_fpack_struct_EQ, 0, Diags);
   Opts.MaxTypeAlign = getLastArgIntValue(Args, OPT_fmax_type_align_EQ, 0, Diags);
-  Opts.AlignDouble = Args.hasArg(OPT_malign_double);
+  Opts.AlignDouble = T.getArch() == llvm::Triple::x86 && Args.hasArg(OPT_malign_double);
   Opts.LongDoubleSize = Args.hasArg(OPT_mlong_double_128)
                             ? 128
                             : Args.hasArg(OPT_mlong_double_64) ? 64 : 0;
   Opts.PPCIEEELongDouble = Args.hasArg(OPT_mabi_EQ_ieeelongdouble);
   Opts.PICLevel = getLastArgIntValue(Args, OPT_pic_level, 0, Diags);
   Opts.ROPI = Args.hasArg(OPT_fropi);
   Opts.RWPI = Args.hasArg(OPT_frwpi);
   Opts.PIE = Args.hasArg(OPT_pic_is_pie);