I meet a problem about code128

If the text structure is like this:

C(n)A(1)B(n) OR C(n)B(1)A(n)

the encoding will calculate incorrectly.

>
```

JsBarcode("#barcode1", "12345678"+"\x08" + "s");
```

the barcode cannot be scan correctly.

I think the problem is in auto.js :

the matched character should not be replaced ,if before it is CODEC。
because you haven't process the condition that set is SET_C , so the CODEA character will be treated as CODEC

Start in auto.js and reproduce the reported JsBarcode example using the Code 128 input containing "\x08" and "s". Compare the generated encoding for the C(n)A(1)B(n) and C(n)B(1)A(n) structures; the issue is done when the resulting barcode encodes and scans correctly.

Written by the indexing model from the issue text.