docs: Add docstrings to public packages (#360)

* Add docstring to the aggregation package
* Add docstring to the torchjd package

Author: ValerianRey

src/torchjd/__init__.py

@@ -1 +1,7 @@
+"""
+This package enable Jacobian descent, through the `backward` and `mtl_backward` functions, which
+are meant to replace the call to `torch.backward` or `loss.backward` in gradient descent. To combine
+the information of the Jacobian, an aggregator from the `aggregation` package has to be used.
+"""
+
 from ._autojac import backward, mtl_backward

src/torchjd/aggregation/__init__.py

@@ -1,3 +1,9 @@
+"""
+This package provides implementations of several popular aggregation strategies from the literature.
+The role of these aggregators is to combine a matrix (e.g. the Jacobian matrix) into a single
+vector, generally to be used like a gradient.
+"""
+
 from ._aggregator_bases import Aggregator
 from ._aligned_mtl import AlignedMTL
 from ._config import ConFIG