fix Ut8Formatter when formatting floating-point whole number

[adienakhmad]

A fix for Utf8Formatter mistakenly removing zeros when formatting a floating-point whole number.

The existing logic would format floating-point value of 10 as "1", removing the zero.
To fix this, the removal of trailing zeros and the decimal separator should be handled separately.

[JunaMeinhold]

Hi, thanks for your pull requests and your work,

The code looks good overall, but I would suggest one change. Instead of checking the separator character directly, it’s safer to store the buffer pointer before writing the separator and later compare the buffer pointer to that stored value. This approach is faster and avoids issues with multi-byte UTF-8 characters, such as some full-width punctuation used in Chinese or other locales.

Full-width period 。 (U+3002)
Full-width comma ， (U+FF0C)
Other full-width numeric punctuation
These characters are multi-byte in UTF-8 (e.g., ， is three bytes: 0xEF 0xBC 0x8C).

byte* beforeSeparator = buffer; 
buffer += ConvertUtf16ToUtf8(format.CurrencyDecimalSeparator, buffer, (int)(end - buffer)); 
byte* afterSeparator = buffer; 

// ... write fraction digits ...

while (buffer != afterSeparator && *(buffer - 1) == '0') { buffer--; } 

if (buffer == afterSeparator) { buffer = beforeSeparator; }

[adienakhmad]

I've updated the code as suggested. I've updated the test as well to use a custom culture as none of the cultures listed in CultureInfo.GetCultures(CultureTypes.AllCultures) uses 3-bytes UTF-8 character as the separator.