There appears to be confusion on whether or not Pleroma can handle this particular
sin of JSON-LD.  It can, and we will add regression tests to prove that it can handle
this particular sin.  Which, by the way, this is actually not a "feature" of JSON-LD,
but whatever.

This is not what with_muted is for per documentation and it was agreed
on irc that this behavior doesn't make sense.

# Conflicts:
#	CHANGELOG.md
#	lib/pleroma/notification.ex
#	lib/pleroma/user.ex
#	lib/pleroma/user/info.ex
#	lib/pleroma/web/admin_api/admin_api_controller.ex
#	lib/pleroma/web/ostatus/handlers/follow_handler.ex
#	lib/pleroma/web/ostatus/ostatus.ex
#	lib/pleroma/web/salmon/salmon.ex
#	lib/pleroma/web/websub/websub.ex
#	test/web/admin_api/admin_api_controller_test.exs
#	test/web/federator_test.exs
#	test/web/mastodon_api/controllers/conversation_controller_test.exs
#	test/web/ostatus/ostatus_controller_test.exs
#	test/web/ostatus/ostatus_test.exs
#	test/web/salmon/salmon_test.exs
#	test/web/websub/websub_test.exs

Misc. fixes / improvements.

While the case tested (`lain@ple`) gives wrong ordering due to
using only trigram to order, it almost never happens in reality. In reality
it would be either `lain` (in which case it's fine to list the user with `lain`
in display name first), or `@lain@pleroma.soykaf.com`/`lain@pleroma.soykaf.com` (which is
handled fine as well)

- Remove trigram as it tends to rank garbage results highly, resulting
in it prioritized above fts, which gives actually decent results. ACKed
by kaniini and lain on irc.
- Remove a test for handling misspelled requests, since we no longer have
trigram
- Remove a test for searching users with `nil` display names, because it
is unrealistic, we don't accept usernames that are not >1 char strings
- Make rank boosting for followers/followees sane again, previous values
resulted in garbage matches getting on top just because the users are
followers/followees

Turns out you can't actually find the user with this.