no-copy semantics for for loops

Note that only the index variable requires a stack allocation, and the
memcpy for the element is gone.

```zig
export fn entry() void {
    var buf: [10]i32 = undefined;
    for (buf) |x| {}
}
```

```llvm
define void @entry() #2 !dbg !35 {
Entry:
  %buf = alloca [10 x i32], align 4
  %i = alloca i64, align 8
  %0 = bitcast [10 x i32]* %buf to i8*, !dbg !47
  call void @llvm.memset.p0i8.i64(i8* align 4 %0, i8 -86, i64 40, i1 false), !dbg !47
  call void @llvm.dbg.declare(metadata [10 x i32]* %buf, metadata !39, metadata !DIExpression()), !dbg !47
  store i64 0, i64* %i, align 8, !dbg !48
  call void @llvm.dbg.declare(metadata i64* %i, metadata !45, metadata !DIExpression()), !dbg !48
  br label %ForCond, !dbg !48

ForCond:                                          ; preds = %ForBody, %Entry
  %1 = load i64, i64* %i, align 8, !dbg !48
  %2 = icmp ult i64 %1, 10, !dbg !48
  br i1 %2, label %ForBody, label %ForEnd, !dbg !48

ForBody:                                          ; preds = %ForCond
  %3 = getelementptr inbounds [10 x i32], [10 x i32]* %buf, i64 0, i64 %1, !dbg !48
  call void @llvm.dbg.declare(metadata i32* %3, metadata !46, metadata !DIExpression()), !dbg !49
  %4 = add nuw i64 %1, 1, !dbg !48
  store i64 %4, i64* %i, align 8, !dbg !48
  br label %ForCond, !dbg !48

ForEnd:                                           ; preds = %ForCond
  ret void, !dbg !50
}
```

1 files changed, 0 insertions, 2 deletions

diff --git a/src/codegen.cpp b/src/codegen.cpp
index 84d5777530..eecc2239db 100644
--- a/src/codegen.cpp
+++ b/src/codegen.cpp
@@ -5539,8 +5539,6 @@ static LLVMValueRef ir_render_instruction(CodeGen *g, IrExecutable *executable,
         case IrInstructionIdInvalid:
         case IrInstructionIdConst:
         case IrInstructionIdTypeOf:
-        case IrInstructionIdToPtrType:
-        case IrInstructionIdPtrTypeChild:
         case IrInstructionIdFieldPtr:
         case IrInstructionIdSetCold:
         case IrInstructionIdSetRuntimeSafety: