Add unified CEBRA encoder: pytorch implementation

[CeliaBenquet]

This PR adds a PyTorch implementation of a unified CEBRA encoder, which is composed of:

• A new sampling scheme that samples across all sessions so that they can be aligned on the neuron axis to train a single encoder.
• A unified Dataset and Loader, adapted to the new sampling scheme.
• A unified Solver that considers multiple sessions to be aligned at inference.
• A new masked modeling training option, with different types of masking.

🚧 A preprint is pending "Unified CEBRA Encoders for Integrating Neural Recordings via Behavioral Alignment" by Célia Benquet, Hossein Mirzaei, Steffen Schneider, Mackenzie W. Mathis.

💻 A DEMO Notebook is available at: https://cebra.ai/docs/demos.html

[MMathisLab]

thanks @CeliaBenquet ! I went through and left comments for disucssion

[stes]

Looks good overall; left some comments!

• Implementation of the Mixin class for the masking: If I understood correctly, the only change is that this apply_mask function is applied after loading a batch. This seems to be a change that could be minimally invasively applied not in the dataset, but actually in the data loader. Is there a good case why the datasets themselves need to be modified?
• Discussion on where to place the decoders: currently in cebra.models.decoders; are the decoders useful as "standalone" models? where are they currently used? based on that we could determine if we move them e.g. as standalone to integrations
• see other comments; mostly on class design, removing duplicated code, etc.

[MMathisLab]

lgtm! Just the one comment on kwargs seems critical to decide

[stes]

Remaining issues are:

• more tests (e.g. for changes to model, adaptations of sklearn API, ...)
• improved class design for masking
• improved class design for solver hierarchy

However, this can be done in follow-up contributions, this can be merged now. Attempt to clean up the commit message for sqashing:

* start tests

* remove print statements

* first passing test

* move functionality to base file in solver and separate in functions

* add test_select_model for multisession

* remove float16

* Improve modularity remove duplicate code and todos

* Add tests to solver

* Fix save/load

* Fix extra docs errors

* Add review updates

* apply ruff auto-fixes

* fix linting errors

* Run isort, ruff, yapf

* Fix gaussian mixture dataset import

* Fix all tests but xcebra tests

* Fix pytorch API usage example

* Make xCEBRA compatible with the batched inference & padding in solver

* Add some tests on transform() with xCEBRA

* Add some docstrings and typings and clean unnecessary changes

* Implement review comments

* Fix sklearn test

* Initial pass at integrating unifiedCEBRA

* Add name in NOTE

* Implement reviews on tests and typing

* Fix import errors

* Add select_model to aux solvers

* Fix tests

* Add mask tests

* Fix docs error

* Remove masking init()

* Remove shuffled neurons in unified dataset

* Remove extra datasets

* Add tests on the private functions in base solver

* Update tests and duplicate code based on review

* Fix quantized_embedding_norm undefined when `normalize=False` (#249)

* Fix tests

* Adapt unified code to get_model method

* Update mask.py

add headers to new files

* Update masking.py

- header

* Update test_data_masking.py

- header

* Implement review comments and fix typos

* Fix docs errors

* Remove np.int typing error

* Fix docstring warning

* Fix indentation docstrings

* Implement review comments

* Fix circular import and abstract method

* Add maskedmixin to __all__

* Implement extra review comments

* Change masking kwargs as tuple and not dict in sklearn impl

* Add integrations/decoders.py

* Fix typo

* minor simplification in solver

---------

Note, some comments in this PR overlap with
https://github.com/AdaptiveMotorControlLab/CEBRA/pull/168
and
https://github.com/AdaptiveMotorControlLab/CEBRA/pull/225
which were developed in parallel.