Fix writing multi-byte characters to tar archive

[TadCordle]

Fixes #626. Looks like we were using string length instead of byte array length.

[loosebazooka]

Can we use codePointCount? Not saying we should, just curious.

[briandealwis]

addEntry() seems to be called by toTarballBlob() to add what's described as a a blob. Why don't we just pass the file contents as byte[] — or better yet as ByteSource — and let the caller deal with the encoding?

If we used ByteSource then we could get rid of the other addEntry(TarArchiveEntry) and use MoreFiles.asByteSource(filePath).

[TadCordle]

So it looks like the main problem has been fixed, but using multibyte characters in tar archive entry header names on windows might still be sketchy (e.g. adding non-ascii filenames might not work). Multibyte file contents should work on all platforms, though. I'll keep looking into it, but feel free to review this now so we can get the fix out.

[coollog]

@briandealwis Blob is essentially our own minimal interface that is essentially a ByteSource, but you make a good point - can you file a separate issue for that since it's not directly related to this fix?

[coollog]

@TadCordle oh we should add a CHANGELOG entry for this

[chanseokoh]

Cool, just in case, have you tested dockerBuild and buildTar?

[TadCordle]

Yeah, tested both with the same args used in the original issue.