Sparql\Client — fallback to autodetection of content-type

[indeyets]

Currently, EasyRdf trusts server to give us a Content-type header corresponding to the body of request. That's good and logical, but some servers are not either good or logical :-D

We should switch to the following approach:

1. If we recognize content-type, then try to parse document according to it
2. if content-type is unrecognized or parser threw EasyRdf\Parser\Exception then use EasyRdf\Format::guessFormat to guess the format
3. If guessing succeeds and guess is different from original format — try parser for the new format
4. if guess fails or gives same format or if new parser fails — throw EasyRdf\Exception