This release has numerous bug and performance fixes:
https://github.com/puma/puma/releases

Noteable changes:

* nakayoshi_fork has been removed (https://github.com/puma/puma/issues/2258).
With compaction disabled, it appears not to provide much gain.

* https://github.com/puma/puma/pull/3035 dropped the use of the `time`
gem in Puma, but our Puma JSON-formatter needs it.

Changelog: changed

Changelog: removed

Puma v6.0 dropped `nakayoshi_fork` in
https://github.com/puma/puma/issues/2925 since "it has very minimal
gain and can trigger nasty bugs in c-extensions due to usage of
GC.compact." If the call to `nakayoshi_fork` is made with Puma 6, the
Web server will crash and fail to start up.

To avoid this, conditionally enable it only if Puma v5 is loaded.

Changelog: fixed

This will be used soon in order to produce
heap dumps when a worker was terminated due
to high memory use.

This was merely testing for truthiness, but any string
is truthy, including "false" or "0".

This caused the worker killer to not start in charts
since here we set this to "false".

Also removed the double negatives.

Changelog: fixed

This code was needed to perform Puma 5 upgrade without syncing multiple
MRs. We update to Puma 5, so this code is no longer needed.

https://github.com/puma/puma/blob/master/5.0-Upgrade.md#nakayoshi_fork
Enabled by default. ENV var for better control.

The changes which required to use our own Puma fork now in the upstream,
so we need to update to it for the future support.

Overwrite ::SnowplowTracker::LOGGER.formatter for Puma(not Unicorn)
Add timestamp to Puma::Events::PidFormatter
Save both of them in JSON format

Currently puma logs are emitted to a different location than rails logs.
In development environments, having a single stream of logs for puma
and rails simplifies things and ensures that developers will see puma
errors, instead of having them emitted to a file that is not usually
checked.

This allows to export /metrics
directly from the master process
instead of exporting from worker process.

PumaWorkerKiller is used for periodically checking and killing
workers (the biggest one) if overall memory reaches specified
limit. This metric allows us to watch number of killed workers.

This allows us (and others) to test drive Puma without it affecting all
users. Puma can be enabled by setting the environment variable
"EXPERIMENTAL_PUMA" to a non empty value.