…ument

The bug really lies in `ref`'s CString type (the `set()`
function specifically), but this is a good test case to
ensure the desired behavior through node-ffi.

See #169.

No strlcpy() apparently :\

Fixes #195.

It's still failing due to #239.
The fix probably won't be backported, so let's just
call v3 a no-go.

It's more low-level than Cocoa, and contains NSException directly.

This isn't a great way to do this.
We need to set up better bechmarks.

For consistency

Except for on Windows where going native makes it return
0 every time for some reason I can't quite figure out.
Leaving the previous logic works though.

Fixes #273.
Closes #284.
Closes #314.

Squashed commit of the following:

commit d13bbe0
Author: Lee, SungUk <sulee@ea.com>
Date:   Fri Feb 19 14:31:21 2016 +0900

    Add `node_ffi_errno` function and recover old `lib/errno.js` code

    V8 native of the some platform(like Windows) will clear errno,
    `ffi.errno()` always return `0`.

    1. now wrap function is just function pointer
    2. windows system must require to use direct ffi binding.
       If not use ffi foreign function binding, `v8` or `uv` will clear
       `errno` value.

commit d27d9aa
Author: Lee, SungUk <sulee@ea.com>
Date:   Thu Feb 18 22:04:12 2016 +0900

    Move `errno` method implement to C++ side

    `errno` value depend on the platform and libc library.
    For example, `glibc` and `musl` differ `errno.h`

    So, Sometimes `lib/errno.js` implement cannot be use.

    Reference:
    - http://sourceware.org/git/?p=glibc.git;a=blob;f=include/errno.h;h=7df41dfc3#l35
    - http://git.musl-libc.org/cgit/musl/tree/include/errno.h?id=2d0290de#n13

Seems to be working as expected

Causing builds to fail on MacOS Sierra.
Fixes #337.

Drop all older versions