Support for post-processors and change in default accumulation behaviour

[gmertes]

The current default behaviour for accumulation variables is to accumulate them at every step with the value of the previous step, starting from the beginning of the forecast.

We are changing that behaviour and will no longer accumulate from the beginning by default. If the user wants this behaviour, it needs to be explicitly turned on in the config.

This change will go together with a new post-processor feature, that allows for extendable post-processors. The accumulation operation will be a post-processor.

The proposed config format is:

post_processors:
  - accumulate
  - do_something_else
  - ...

(exact naming is not determined yet, open to feedback)

The default accumulation behaviour will be rolled out in steps, so users have time to adjust:

1. Existing users will not have the post_processors entry in their config. For those users, we still accumulate by default but show a warning that this behaviour will change in the future.
2. Users can explicitly turn accumulation on or off by specifying the post_processors entry in the config, with either an empty list or accumulate in the list.
3. Finally and after some time, for users that still do not have post_processors in their config, we change the default behaviour to not accumulate by default anymore.