Multitask ASE interface

[laserkelvin]

This PR closes #217 by providing an extended interface for mapping multitask models to ase workflows.

The main changes include:

• Introduces the matsciml.interfaces.ase.multitask.AbstractStrategy class, whose children allows us to implement ways to aggregate the outputs of multitasks.
• Implements the sole concrete class (for now), AverageTasks, which does a mean aggregation across the different dataset heads to compute energies/forces. Essentially, if you had ForceRegressionTask trained on multiple datasets, we outputs from all.
• Implements a custom ase_calculate method for MultiTaskModule, which effectively streamlines inference using multitask modules in scenarios like as an ASE calculator. It's named specifically for ase, but in the future we can use it as a dedicated inference method.

The AbstractStrategy should be flexible enough to do other potentially more intelligent aggregations, e.g. a weighted average, MoE approaches, etc.

[laserkelvin]

Ergh, issue with one of the tests...

[laserkelvin]

@melo-gonzo just wanted you to take a look at the example before I merge, and if you wanted to clarify what you meant in your question?

[melo-gonzo]

@laserkelvin thanks for the example! I am just curious why you made the change in that line. Feel free to merge ✅