Port distributed training support from existing PR

I'm opening this PR to keep track of the work needed to port the content of the #996 PR to the main branch.

The idea is to split that PR (which is huge and based on a quite old version of the codebase) and, starting from the current state of the main branch, port its main elements in smaller PRs.
I'll keep this issue updated as I work on this.

Many changes are not strictly related to supporting distributed training but may benefit Avalanche in general.

1. I'm starting with porting the modernized object detection/segmentation dataset, strategies, and metrics. I'll also port the generalized batch collate functionality.

---
# Changes in Distributed Training PR #996:

Legend:
- :black_square_button: Not ported
- :hourglass: Work in progress
- :speech_balloon: PR opened, discussion in progress
- :heavy_check_mark: Merged into main branch

## Base elements
- :heavy_check_mark: DistributedHelper implementation (#1370)
- :black_square_button: Distributed value, object, batch, model, tensor, ...
- :heavy_check_mark: Distributed consistency (hashers) (#1370)
- :black_square_button: Distributed training example (and runner script)

## Strategy e plugins
- :heavy_check_mark: New `supports_distributed` plugin field (#1370)
- :heavy_check_mark: New `_distributed_check` strategy field and related `_check_distributed_training_compatibility()` check (#1370)
- :black_square_button: New `wrap_distributed_model` strategy lifecycle method. Called from `..._observation.py`
- :heavy_check_mark: `_obtain_common_dataloader_parameters` strategy method (unrelated to distributed training) (#1370)
- :black_square_button: Strategy support superclasses
- :black_square_button: Various plugin adaptations for distributed training (LwF, CWR, ...)
- :heavy_check_mark: AR1: modernize to use `_obtain_common_dataloader_parameters` (unrelated to distributed training) (#1370)
- :black_square_button: Strategy templates: wrap various lifecycle methods to allow for seamless support of distributed training
- Implementations should now be in `_backward()`, `_forward()`, ... while wrapping happens in `backward`, `forward`. Wrapper methods should be *final*, but Python is not strict on this (flexibility).

## Models
- :heavy_check_mark: Fixed device issues with dynamic models (unrelated to distributed training) (#1370)
- :heavy_check_mark: In `avalanche_forward`, generalize using `is_multi_task_module` to consider DDP wrapping (#1370)

## Detection
- :heavy_check_mark: Detection scenario modernization (#1333)
- :heavy_check_mark: Detection template (incl. Naive) modernization (#1333)
- :heavy_check_mark: Updated detection example (#1333)
- :heavy_check_mark: Detection dataset based on new dataset creation procedure (#1333)
- :hourglass: Collate generalization

## Data Loader
- :heavy_check_mark: Use DistributedHelper, remove mock _DistributedHelper (#1370)
- :heavy_check_mark: Various fixes to address drop_last, shuffle, etcetera

## Loggers and metrics
- :heavy_check_mark: Disable loggers creation for non-main processes (#1370)
- :heavy_check_mark: Default logger: pass 'default' instead of loggers list (#1370)
- :heavy_check_mark: Strategies constructor: allow strategies to accept a factory for the `evaluator` constructor parameter (`evaluator=default_evaluator()` -> `evaluator=default_evaluator`). (#1370)
- :heavy_check_mark: All strategy classes: change the default `evaluator` parameter value to use a factory. (#1370)

## Unit tests
- :black_square_button: Called in both environment-update and unit-test actions
- :heavy_check_mark: Unit test runner: run_dist_tests.py and related utils (#1370)
- End-to-end test script

## Typing
- :heavy_check_mark: Various typing fixes/integrations in AvalancheDataset and FlatData (#1333)
- Mostly to improve the programming experience of VSCode users