Better handling of invalid UTF8 strings.

[ExtremeMan]

As it appears iconv could fail and set errno to EINVAL even if ICONV_SET_DISCARD_ILSEQ is set [1]. In that case xctool should skip invalid UTF8 bytes [2], replace every invalid sequence of bytes with a replacement character [3].

In this PR StringFromDispatchDataWithBrokenUTF8Encoding was updated to handle use cases described above. If errno is EINVAL then function skips all bytes until possible start of new sequence is detected. Then iconv is used again to read them. At the end new string is composed from valid UTF8 substring and returned to a caller.

Fixes #662.

Ref:
[1] https://developer.apple.com/library/mac/documentation/Darwin/Reference/ManPages/man3/iconv.3.html
[2] https://en.wikipedia.org/wiki/UTF-8
[3] https://en.wikipedia.org/wiki/Specials_(Unicode_block)#Replacement_character

[ExtremeMan]

@yiding, addressed your feedback.

[ExtremeMan]

Using initWithBytesNoCopy:length:encoding:freeWhenDone: as suggested for efficiency.