Feature: Pulsating Sphere Datasets

[JakobEliasWagner]

Description

This PR introduces two new datasets: ConstBoundaryDataset and InverseConstBoundaryDataset. Both datasets are designed to work with pulsating sphere data, with constant boundary conditions, and leverage ADIOS2 for data reading.

##Key Features:

• ConstBoundaryDataset:

  1. Loads a pulsating sphere dataset with constant boundaries.
  2. Supports loading pressure data (real and imaginary parts) and geometry from .bp files.
  3. Provides boundary condition samples for top, right, and frequency.
  4. Normalizes the input features and outputs to prepare the data for training.

• InverseConstBoundaryDataset:

  1. Designed for inverse problems where the boundary conditions are constant.
  2. Supports selecting a subset of sensors for training.
  3. Allows the specification of boundary condition points and observation selections.
  4. Offers normalized inputs and outputs to maintain consistent scales.

• Additional Changes

  1. Utilizes adios2 for handling .bp files.
  2. Introduces new normalization and scaling strategies for the datasets.
  3. Added support for custom observation and sensor indexing.

How are the changes tested?

• Unit-tests:
  • Initialization
  • Correct population of tensors.

Checklist

• Documentation
  • All new features include documentation
  • README is updated
• CI/CD passes all pipeline checks
• Post Merge
  • Delete the feature branch (if applicable).
  • Update related issues or project boards.