Unicode normalization combinators

[adithyaov]

Todo:

• Add benchmarks

[adithyaov]

Ignore cabal.project. We'll release unicode-data before merging this.

[harendra-kumar]

@adithyaov can you rebase this, port it on the new unicode-data package, and add benchmarks from unicode-transforms package?

[harendra-kumar]

1. Let's add the normalization routines to the Unicode.Char module. We will use this module for operations on Char type including stream of Char to Char transformations. The Unicode.Stream module should go away. We will move the encode/decode routines in Unicode.Stream to Unicode.Utf8 and Unicode.Latin1.
2. Let's add some basic benchmarks as well so that we can start working on performance optimizations of these modules. Just take an input file and normalize it. We already have some benchmarks in the existing Unicode.Stream benchmarks that take input files using env vars.

Once the normalization tests pass and we have benchmarks in place, we can commit this and do the performance optimizations later.

[harendra-kumar]

call this composeD.

[adithyaov]

This is kind of partially composing. This does not compose a semi-decomposed stream. It shouldn't be used directly.

[harendra-kumar]

Looks good at a high level. I did not check the details.