Handle subnormal properly in close_to_linear

The old code just assumed that if both A and B were close
to zero that B would not dominate A forcing linear
behavior. This is not necessarily the case, so check to
make sure.

Change-Id: I0ff4098ece02abbb49b2ba0ae4f429b5dccb8558
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/705226
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Herb Derby <herb@google.com>
1 file changed