RFC: Use a lossy conversion in ResponseReader::text_utf8

[adamreichold]

Fixes #87

[Shnatsel]

I admit I have no idea what the correct behavior is here. It's possible that refusing to process invalid UTF-8 is the right thing do to, and the lossy conversion used by e.g. ureq is wrong. Are there any standard specifying this?

[sbstp]

It seems like reqwest also uses a lossy conversion https://docs.rs/reqwest/0.11.0/reqwest/struct.Response.html#method.text

[adamreichold]

We also decided for lossy conversion when using encoding_rs as encoding_rs_io defaults to this behaviour. So purely for reasons of consistency, the lossy versions seems preferable.