Fix multibyte bug on utf-8 import

[filkaris]

When importing a csv file, we currently take the first 1024 bytes as a sample via file_get_contents, and try to detect its encoding.

When, however, the file is in utf-8 encoding, there is a chance that the first 1024 bytes split a multibyte character. As a result the encoding is not recognized as utf-8.

This fixes that issue by using php's mb_substr function.

[ralfbecker]

I'd rather use the following:

$sample = mb_substr(file_get_contents($content['file']['tmp_name'], false, null, 0, 2048), 1024);

Please do not assign to $sample twice, as this is flagged as potential mistake in Netbeans.