Compatibility with GHC's JavaScript backend

[clyring]

It's my understanding that the C FFI as used by the bytestring library is not well-supported by the JavaScript backend. To work around this we can add fallback Haskell-only implementations and use these whenever an appropriate cabal flag is set.

See also earlier discussion at #573.

[clyring]

Question: #587 added more reliance on the C FFI because of alignment issues. I assume the unaligned writes themselves are not really a problem, but the patch also moves some static data tables from Haskell code into C code so that we can read multi-byte values from them in Haskell code without worrying about alignment or endian-ness. Will that need a workaround? I hope it's easier for the JS backend to support than passing ByteArray#s and such into general C code.