Use portable c_char instead of i8

[ghost]

Some architectures, such as ARM, use unsigned 8-bit integers to represent C char type.

In order to make the code work on architectures with both signed and unsigned chars, this PR replaces i8 by c_char in places where it corresponds to C char type.

[skade]

As we're here already, we could also modernise the codebase cough and make this function be pub(crate). Also, this is one of the rare cases where I would be okay with accepting a patch that uses from_utf8_unchecked, as I trust the leveldb maintainers to have their error messages be valid utf8 in all cases.

[ghost]

I trust the leveldb maintainers to have their error messages be valid utf8 in all cases

I would not trust them in this, as they are using an ordinary C++ std::string which is not UTF-8 aware. In particular, I was able to generate non-UTF-8 error message with this code on Linux:

use leveldb::{database::Database, options::Options};
use std::{ffi::OsStr, os::unix::ffi::OsStrExt, path::Path};

fn main() {
    let _ = Database::<i32>::open(Path::new(OsStr::from_bytes(b"/\xc3\x28")), Options::new())
        .map_err(|e| println!("{:?}", e));
}

where the path is not valid UTF-8. I've created a separate PR #34 on top of this one that handles non-unicode strings explicitly.

[skade]

That is correct. Thanks for the patch! I left one comment.

[michaelsproul]

I believe this could be closed as #36 achieved the same goal.

[skade]

Closing as suggested. Thanks for the patch nevertheless, @michaelsproul!