Add catchable failure decoders for more encodings:

[mrijkeboer]

This adds the following decoders:

• decodeUtf16LE'
• decodeUtf16BE'
• decodeUtf32LE'
• decodeUtf32BE'

[bos]

These seem like good ideas, but I have two observations.

The names aren't exactly self-documenting :-)

The implementations seem a bit unnatural. Basically if you're round-tripping through unsafePerformIO . try . evaluate, maybe the underlying primitives have the wrong return types?

[mrijkeboer]

The code is based on the already provided decodeUtf8' function. I tried to keep it as consistent with the existing code as possible.