* chore: update dependencies

* update @types/node,tsd, typescript

* remove @typescript-eslint

* more updates

* update uuid

* update CI to node 20,22,24

* chore: migrate to ES6 classes

* add constructor check to test/basic.js

* add teardown to Aedes constructor test

* fix basic.js

* feat: automate benchmark testing

* group averages

* add percentages to report.js

* fix missing counter reset in bombingQoS1

* add README.md and examples

* update bombingQoS1

* remove benchmark examples

* improve benchmarking

* put summary first

This PR migrates Aedes to use the async persistence interface.

It contains the following sub items:

- [X] replace all callback calls to persistence by `.then()` calls
- [X] move all side effects (setting up persistence, attaching event
handlers etc) to a `listen()` method of Aedes
- [X] update `createBroker()` to be async and make it await `listen()`
as well
- [X] update exports so that they produce a warning if people use old
style calling
- [X] update Typescript typing
- [X] update documentation including examples
- [X] add a `migration.md` to help users migrate.
- [X] do some benchmark tests to see performances are kept
- [X] update node engines version to >=20
- [X] updates all dependencies to current except for Tap and
@sinonjs/fake-timers

BREAKING CHANGES: See
[docs](https://github.com/moscajs/aedes/blob/main/docs/MIGRATION.md)

This PR adds Github actions to automate benchmark running on PR requests

It will create a [Github Job
Summary](https://github.blog/news-insights/product-news/supercharging-github-actions-with-job-summaries/)
of the benchmark and add it to the benchmark action (see
https://github.com/moscajs/aedes/actions/runs/16239498783#summary-45853979456
)

The benchmark steps run serialized in the hope that performance on a
single runner does not vary too much.
I have been building testing this using
seriousme#1 which also shows how it
operates.

It also shows that variation between results is inevitable, even when
running on the same runner and the Aedes code is identical !!

In seriousme#1 I was able to add the
comment automatically to the PR using
`.github/actions/sticky-pr-comment/action.yml` however if you send a PR
from a fork the GITHUB_TOKEN that runs the actions by default (event:
[pull_request](https://docs.github.com/en/actions/reference/events-that-trigger-workflows#pull_request)
) only has `read` access on the PR for security reasons (see [Pull
Request events for forked
repositories](https://docs.github.com/en/actions/reference/events-that-trigger-workflows#pull-request-events-for-forked-repositories))

This can be fixed by attaching to the event:
[pull_request_target](https://docs.github.com/en/actions/reference/events-that-trigger-workflows#pull_request_target),
however Github notes explicitly: *Avoid using this event if you need to
build or run code from the pull request.*

So for now I used Github Job Summary instead.

Kind regards,
Hans

---------

Co-authored-by: Daniel Lando <daniel.sorridi@gmail.com>

This PR migrates Aedes to ESM.
I followed Sindre Sorhus' checklist mentioned by Daniel in #1038:

- [X] Add "type": "module" to your package.json 
- [X] Replace "main": "index.js" with "exports": "./index.js" in your
package.json.
- [X] Update the "engines" field in package.json to Node.js 20: "node":
">=20".
- [X] Remove 'use strict'; from all JavaScript files.
- [X] Replace all require()/module.export with import/export.
- [X] Use only full relative file paths for imports: import x from '.';
→ import x from './index.js';.
- [X] If you have a TypeScript type definition (for example,
index.d.ts), update it to use ESM imports/exports.
- [X] Use the [node:
protocol](https://nodejs.org/api/esm.html#esm_node_imports) for Node.js
built-in imports.

And added the following items:
- [X] remove `proxyquire` as it does not work with ESM (Proxy-require,
the name says it all)
- [X] monkey patch neostandard in eslint.config.js as it does not
support import assertions yet (I opened a ticket with them, but this
works)
- [X] add `test/require.cjs` which checks if `require("aedes")` still
works for those that need it.

Enjoy!
Hans
For reference: I used `ts2esm` with a simple `jsconfig.json` to do most
of the heavy lifting.

---------

Co-authored-by: Daniel Lando <daniel.sorridi@gmail.com>

This PR updates all dependencies to their latest versions:
 - @sinonjs/fake-timers   ^11.2.2  →   ^14.0.0
 - @types/node           ^24.0.12  →  ^24.0.14
 - eslint                 ^9.30.1  →   ^9.31.0
 - mqemitter               ^7.0.0  →    ^7.1.0
 - release-it             ^19.0.3  →   ^19.0.4
 - tap                   ^16.3.10  →   ^21.1.0

It removes the following dev dependencies that were no longer used:
- typescript
- snazzy

The upgrade of tap required:
- a new .taprc as the contents were not relevant anymore for v21
- adjustment of package.json as e.g. the Tap -J flag is now default
behavior
- fixing of a few tests in `client-pub-sub.js ` that incorrectly passed
too many arguments to t.pass()
- adjustment of CI.yml because the location of the coverage folder was
changed by Tap

It also removes a console.log() from `test/connect.js` which was a
leftover from debugging.
 
As a bonus I modified CI.yml slightly so that the first job of the
matrix will always upload coverage and we do not need to update the
nodeJS version in this specific "if" anymore.

Kind regards,
Hans

This PR replaces Tap by node:test

Tasks:
-  [x] auth.js
-  [x] basic.js
-  [x] bridge.js
-  [x] client-pub-sub.js
-  [x] close_socket_by_other_party.js
-  [x] connect.js  (with TODO's)
-  [x] events.js
-  [x] keep-alive.js
-  [x] meta.js (with 1 TODO)
-  [x] not-blocking.js
-  [x] qos1.js
-  [x] qos2.js
-  [x] regr-21.js
-  [x] require.cjs
-  [x] retain.js (with 1 TODO)
-  [x] topics.js
-  [x] will.js
-  [x] update package.json scripts
-  [x] update package.json devDependencies
-  [x] update 'coverage' directory in ci.yml

The following devDependencies have been replaced by NodeJS native
functionality
 - @sinonjs/fake-timers
 - concat-stream
 - duplexify
 - mqtt-connection
 - tap

It adds `c8` as devDependency to facilitate test coverage.

A bug in the benchmarking report triggered a fail if the benchmark of
the branch to merge scored more than the threshold over the main branch
(ie. the merge branch is better), where it should fail if the score is
more than the threshold below the main branch.

This PR aims to fix that, and also improves reporting in case of a fail.

Kind regards,
Hans

…release-it.json (#1052)