[ENH] Implement Disjoint-CNN deep learning for classification and regression and adding mutlvariate specificity tag

[hadifawaz1999]

Describe the feature or idea you want to propose

Having specific multivariate deep learning models is a good thing, i like to have Disjoint-CNN, Monash's multivariate CNN model [1] into aeon, its implementation is already quite clear and easy and is in tensorflow here

[1] Foumani, Seyed Navid Mohammadi, Chang Wei Tan, and Mahsa Salehi. "Disjoint-cnn for multivariate time series classification." 2021 International Conference on Data Mining Workshops (ICDMW). IEEE, 2021.

Describe your proposed solution

• The DisjointCNNNetwork class should be added to the network class but parametrized to number of layers, number of filters, kernel size per layer. HINT: look on how already FCN ResNet and all other CNN networks are already implemented

• Adding The DisjointCNNClassifier and DisjointCNNRegressor classes to deep classification/regression models

• Add a tag for the networks module, deep classification and regression modules called "multivariate_univariate_duality" or something like that:

• 1. if this flag is True, means that it works fine with both
• 2. if its false, it means it was proposed for multivariate, it can technically work on univariate but doesnt make sense, case example: DisjointCNN

Describe alternatives you've considered, if relevant

No response

Additional context

i already have a code ready for it, adding this issue to remember the tagging thing, assigning myself