type inference regression

[Mark-Simulacrum]

• root: varkor/metasub: start v. end

I usually deem these as not too interesting but this one felt a bit out of the norm -- it looks like &Foo.to_string_lossy() used to (deref?) coerce to &str but not sure. I think it's fine to just close this if we don't want to do deeper investigation here.

cc @varkor

[varkor]

I think this was observed in #64856 (comment) and it was agreed in FCP to accept this regression: #64856 (comment). Thanks for noting it, though!