Prototype implementation of filter selection

This rewrites the code used previously in the heuristics to make it easier to
debug and introduces the 'methodical' method, which is intended to be an
expensive but reliable way of reducing image size.

The code in this commit does not work; the 'methodical' test for success does
not take account of data buffered inside zlib and, anyway, it changes the
results of pngtest so that the test fails.  This commit is just a checkpoint of
the current state; another commit will temporarily disable the 'methodical'
code.

Signed-off-by: John Bowler <jbowler@acm.org>

pngmem.c

@@ -87,13 +87,17 @@ png_malloc_base,(png_const_structrp png_ptr, png_alloc_size_t size),
    if (size > 0 && size <= PNG_SIZE_MAX)
 #endif
    {
+      png_voidp result;
+
 #ifdef PNG_USER_MEM_SUPPORTED
       if (png_ptr != NULL && png_ptr->malloc_fn != NULL)
-         return png_ptr->malloc_fn(png_constcast(png_structrp,png_ptr), size);
+         result = png_ptr->malloc_fn(png_constcast(png_structrp,png_ptr), size);
 
       else
 #endif
-         return malloc((size_t)size); /* checked for truncation above */
+         result = malloc((size_t)size); /* checked for truncation above */
+
+      return result;
    }
 
    else