It was sending MESSAGE_ERROR, which was inappropriate. Send an empty batch
for now (this is at parity with the in-memory implementation).

Reported by @mogad0n. If a user had both operator privileges and
channel owner privileges, the CS TRANSFER would proceed as though
unprivileged, requiring acceptance by the receiving user. Fix this
to not require acceptance.

Reported by @tacerus. This was most likely introduced by 8b2f6de3e.

Add a config option to suppress LUSERS

See #1676

Ensure the pagination window is full by making sure that every history item
gets a replay line in CHATHISTORY output, even TAGMSG.

This reverts commit 5bbee02fe6.

Fix various pagination issues with CHATHISTORY; also undo #491
(msgid munging).

Fixes #1787

Partially fixes #1679, not sure whether to police the relaymsg config or not

Warn if the user sent, e.g., /QUOTE /SAMODE instead of /QUOTE SAMODE

RPL_WHOISACTUALLY should display some arbitrarily chosen IP address and
hostname.

* Add email-based password reset

Fixes #734

* rename SETPASS to RESETPASS

* review fixes

* abuse mitigations

* SENDPASS and RESETPASS should both touch the client login throttle
* Produce a logline and a sno on SENDPASS (since it actually sends an email)

* don't re-retrieve the settings value

* add email confirmation for NS SET EMAIL

* smtp: if require-tls is disabled, don't validate server cert

* review fixes

* remove cooldown for NS SET EMAIL

If you accidentally set the wrong address, the cooldown would prevent you
from fixing your mistake. Since we touch the registration throttle anyway,
this shouldn't present more of an abuse concern than registration itself.

Advertising SCRAM-SHA-256 breaks irccloud, which doesn't fall back to PLAIN
if it sees SCRAM advertised but SCRAM then fails (as is the case for any
account password hashed on Ergo 2.7 or lower).

Leave a config option for irctest to enable it in the controller.

Refs:

* other instances in the codebase
* https://defs.ircdocs.horse/defs/numerics.html#err-needmoreparams-461
* https://modern.ircdocs.horse/#errneedmoreparams-461

For consistency with RFC2812, Bahamut, Hybrid, Insp, Plexus4, Unreal.
https://datatracker.ietf.org/doc/html/rfc2812#section-3.2.8

At the expense of consistency with chary/solanum, irc2, and ircu2.

This is already implemented, but TARGMAX=KICK:1 says it isn't.

Instead, let's advertise that indefinitely many targets are allowed.
Refs:

* https://defs.ircdocs.horse/defs/isupport.html#targmax
* https://github.com/ircdocs/modern-irc/pull/112

This should be disallowed; `MODE #keytest +k :` should just be an error.

"LOGOUT ALL" logs out all clients for a user.

Fixes #1712.

reported by @Mikaela

We can check once during initialization of the SASL session, e.g.
on receiving `AUTHENTICATE PLAIN` or `AUTHENTICATE EXTERNAL`

Fixes #1762

See: https://github.com/ircv3/ircv3-specifications/pull/466

A default channel mode of +i would block channel creation; fix this by treating
initial joins as SAJOINs.

Note that it's nontrivial to detect initial join in (*Channel).Join, because
having 0 members does not necessarily indicate a new channel.

Allow SAJOIN or implicit always-on join to override channels.operator-only-creation

RENAME (channel rename) that was a simple case change (e.g.
renaming #chan to #CHAN) would delete the channel :-|

This is one of the two cases of [RFC 2812 kicks](https://datatracker.ietf.org/doc/html/rfc2812#section-3.2.8):
even when there are multiple user targets, the RFC (and Unreal
and Inspircd and probably others) allows a single channel name.

Fixes #1740

See #1659

Fix error message for privileged NS PASSWD on a nonexistent account

May help clarify TLS configuration issues, possibly others too