Trying to do a two-release migration proved to be very hacky and ugly.
Instead this is preparation for moving defaults first and then merging
the old configuration left in admin's config file at runtime while still
emitting a warning. This hopefully will be a lot nicer and less
complicated.
Elixir 1.19 started emitting warnings about our use of:
config :logger, :console and
config :logger, backends: []
which got split in 1.15 to :default_handler, :default_formatter, and
LoggerBackends.add(backend) respectively.
On Pleroma startup it is now necessary to add the additional logger
backends at runtime (done with configure_logger() in application.ex).
New config key :pleroma, :logger_backends: [] has replaced the older
backends array. The logger_backends package is now used to facilitate
usage of other logger backends besides the default one (:logger_std_h
from OTP).
Warnings were added to Pleroma startup to warn about the usage of
:logger, backends and additional logic was introduced to make addition
of ExSyslogger to admin's configuration less dumb.
Ref: https://hexdocs.pm/logger/1.15.0/Logger.html#module-backends-and-backwards-compatibility
It's rare that info logs are needed to debug the issue, so I would
suggest setting them to warn in prod by default to make finding the
relevant parts easier and potentially even decrease cpu usage on bigger
instances
Closes#962
