tvg-render: use bufferedReader for 10x performance boost in parsing

I took a profile of tvg-render and found that parsing the file took about
68% of the time.  This was surprising as I would have expected rendering
to take longer.  The profile revealed that most of time was spent in
the filesystem read syscalls.  I've seen this problem before and the
solution in the past has been to use a buffered reader rather than the
file reader directly.  After adding the buffered reader the parse time
went from about 2 ms to about 200 microseconds (10x improvement).

src/tools/render.zig

@@ -114,6 +114,7 @@ pub fn main() !u8 {
     }
 
     // TODO: Render here
+    var buffered_reader = std.io.bufferedReader(source_file.reader());
 
     var image = try tvg.rendering.renderStream(
         allocator,
@@ -129,7 +130,7 @@ pub fn main() !u8 {
         else
             .inherit,
         @enumFromInt(super_scale),
-        source_file.reader(),
+        buffered_reader.reader(),
     );
     defer image.deinit(allocator);