gdiplus: Reimplement metafile loading using gdi32 instead of IPicture.

[Vincent Povirk]

I'd probably be OK with this as-is, but since you asked for feedback:

> @@ -2219,12 +2155,6 @@ GpStatus WINGDIPAPI GdipGetImageBounds(GpImage *image, GpRectF *srcRect,
>          srcRect->Height = (REAL) ((GpBitmap*)image)->height;
>          *srcUnit = UnitPixel;
>      }
> -    else{
> -        srcRect->X = srcRect->Y = 0.0;
> -        srcRect->Width = ipicture_pixel_width(image->picture);
> -        srcRect->Height = ipicture_pixel_height(image->picture);
> -        *srcUnit = UnitPixel;
> -    }
>
>      TRACE("returning (%f, %f) (%f, %f) unit type %d\n", srcRect->X, srcRect->Y,
>            srcRect->Width, srcRect->Height, *srcUnit);

Even for ipicture metafiles, this case wasn't really used before,
except when given an invalid image object (type isn't bitmap or
metafile) where it would likely crash. We should probably return
InvalidParameter here.

Same with the other GdipGetImage* functions where we're removing else clauses.

I am OK with the simplified error handling from IStream_Read.

> +    hr = IStream_Read(stream, buf, emh.nBytes, &size);
> +    if (hr == S_OK && size == emh.nBytes)
> +    {
> +        hemf = SetEnhMetaFileBits(emh.nBytes, buf);
> +        if (hemf)
> +        {
> +            status = GdipCreateMetafileFromEmf(hemf, TRUE, metafile);
> +            if (status != Ok)
> +                DeleteEnhMetaFile(hemf);
> +        }
>      }
>
> +    heap_free(buf);
> +    return status;
> +}

I guess this works, but it's kind of confusing how status gets set
here when IStream_Read or SetEnhMetaFileBits fails. I'd rather it be
set closer to the failure.