Split encoding::types into encoding-types crate

[nagisa]

This allows for creation of alternative non-WHATWG encodings that use the same
interface as encodings defined in this crate without pulling in all the
tables and encodings.

This commit does not introduce any breaking changes; all the types previously defined in
encoding::types are reexported.

Fixes #81.

---

I tried to avoid breaking changes for now, but IMO fn decode being in encoding::types makes little sense; I’d move it to encoding at some point later.

[lifthrasiir]

It looks mostly fine to me, with the following caveats:

• I agree that encoding::types::decode is in the strange location (it was a result of aggressive module split, sigh), but we need to wait until 0.3 to get rid of it. You may want to move it into the crate root and reexport it from encoding::types.
• I guess most (and in fact all, but I haven't tried that) types tests should be actually in a new crate.

[lifthrasiir]

Is this split intentional?

[nagisa]

Yes, ROT13 is defined to only work on basic latin characters (a to z and A to Z). In ASCII table uppercase and lowercase letters are not adjacent, therefore the split.

[lifthrasiir]

I meant that 'a'...'z' | 'A'...'Z' would probably work.

[nagisa]

I guess most (and in fact all, but I haven't tried that) types tests should be actually in a new crate.

The tests depend on encoding::util which is a private module, therefore moving tests without rewriting them to not use index_iter is not possible. Do you still want to move them? I’ll change tests to not use index_iter then.

[lifthrasiir]

@nagisa Ah, that's a legitimate reason (and my suggestion can be incrementally implemented anyway). I have no reason not to merge this, so I went ahead; thank you for the PR!