Merge #604: Clarify the documentation of normalize_s

f6c68ec Clarify the documentation of `normalize_s` (Matt Corallo)

Pull request description:

  I was reading the docs for `normalize_s` and got confused what the point was - it says that libsecp "will only accept" signatures that are normalized, which led me to believe it would refuse to deserialize such signatures. This is untrue, it only refuses to *validate* such signatures.

ACKs for top commit:
  apoelstra:
    ACK f6c68ec

Tree-SHA512: 27ac2f8819638fb889d0fbfb4bc1c07059854b8a15bf2dc2c4140eaf805eb15665fe3d87cadefd56c7b6200ef39818e8d6602e87d7ae7c1d4a4229d4829ea3d0

Author: apoelstra

src/ecdsa/mod.rs

@@ -124,8 +124,8 @@ impl Signature {
     /// changing even the signature itself can be a problem. Such applications
     /// require a "strong signature". It is believed that ECDSA is a strong
     /// signature except for this ambiguity in the sign of s, so to accommodate
-    /// these applications libsecp256k1 will only accept signatures for which
-    /// s is in the lower half of the field range. This eliminates the
+    /// these applications libsecp256k1 considers signatures for which s is in
+    /// the upper half of the field range invalid. This eliminates the
     /// ambiguity.
     ///
     /// However, for some systems, signatures with high s-values are considered