GMT sometime gives garbled Chinese characters

GMT can plot Chinese characters after some simple configurations, but for some Chinese characters, the output is quite garbled.

Steps to reproduce the issue:

1. The attached file [text.txt](https://github.com/GenericMappingTools/gmt/files/4022164/text.txt) is a plain text file, which contains only one record `5 5 中文剑号巨阙`. **The text is GB-2321 encoded, not UTF8**. So you have to open it with "GB-2321" encoding to see the real Chinese characters.
2. Run `gmt pstext text.txt -R0/10/0/10 -JX10c > map.ps`
3. Open the PS file `map.ps` with encoding "GB-2321". At the end of the PS file, you will see
. The first three Chinese characters are correct, and the others are wrong.

With GB-2321 encoding, the hex codes of these Chinese characters are (two bytes for each Chinese character).
```
D6 D0 CE C4 BD A3 BA C5 BE DE E3 DA
```
The hex code of `号` is `BA C5`, and the hex code of `巨` is `BE DE`. However, in the function `psl_fix_utf8`, `C5 BE` (one byte from `号` and one byte from `巨`) is interpreted/fixed as `ž`, resulting in the garbled Chinese characters.