Fix a previously undetected bug in the specification.

The `encode` function was mistakenly expected (by the specification) to produce
an invalid Bech32 string, when given an upper-case human readable part. If this
invalid string is passed to the `decode` function, it fails (unsurprisingly) to
decode.

This change updates the specification so that the `encode` function is
now expected to produce a valid Bech32 string.

lib/bech32/test/Codec/Binary/Bech32Spec.hs

@@ -107,7 +107,7 @@ spec = do
 
         it "hrp lowercased" $ do
             let (Right hrp) = humanReadablePartFromText "HRP"
-            Bech32.encode hrp mempty `shouldBe` Right "hrp1g9xj8m"
+            Bech32.encode hrp mempty `shouldBe` Right "hrp1vhqs52"
 
     describe "Arbitrary ValidBech32String" $