"inline" sender adaptors should forward the `get_forward_progress_guarantee` query

[ericniebler]

the get_forward_progress_guarantee query is not a forwarding query (by design). an adaptor like then, which doesn't effect a sender's forward progress guarantee, should forward that query explicitly.

[ericniebler]

i suppose the function passed to then could actually block, which would be bad. needs thought, and an audit of the adaptors.