vm unpackString does not tolerate invalid unicode

[swdunlop]

Due to how Python handles unicode, vm.unpackString can fail with decoding errors when invalid codepoints are received from Dalvik. For forensic use, this is terrible behavior.

To support this, a wrapper class should be written that preserves this data as a bytestring and presents Python unicode strings with omitted invalid codepoints as a str to prevent dependent functions from throwing exceptions when operating on derived data.

See also how vm unpacks chr's:

https://github.com/swdunlop/AndBug/commit/9ae4bd28ec8bc1f4e0bd60fd988c8931072ffa6c#commitcomment-1426109