Be more resilient to strange codepoints in encoding

[benoit74]

In https://marxists.incn.su/deutsch/archiv/lafargue/1884/xx/appetit.htm (and many others), the - is not the standard -:

<meta http-equiv="content-type" content="text/html; charset=iso-8859&#8211;1" />

While this is mistake from the person who built this website, we could make the scraper resilient to this and automatically transform in proper encoding (iso-8859-1 in example above). Currently the scraper consider that encoding is iso-8859 in example above and ... fails.

For this we need to:

• adapt the regex searching for the charset in document headers
• replace strange codepoints by their ascii equivalent