Don't rely on the system locale for converting to/from bytes.

By default, when java converts Strings to bytes, it uses the
default system locale.  This can be specified by the -Dfile.encoding
option.  If no file encoding is specified, java uses ISO8859_1.

Unfortunately, not all unicode characters can be mapped to
ISO8859_1.  Unmappable characters may be replaced by a byte
within ISO8859_1, which may change the meaning of the String.
This is especially problematic for password strings, and has
been used to compromise the security of passwords in the
past.

Thankfully, Android uses UTF-8 by default, so this bug doesn't
effect Android devices.  However, it's recommended to explicitly
list the character set when converting to/from bytes to
avoid the potential ambiguity.

Change-Id: Iec927e27ed3fc103696c439f6bd3e8779a37ade8