[Hierarchical] Simplify computation of projection matrix using the standard summation matrix

[rshyamsundar]

Issue #, if available:

Description of changes:

We can compute the projection matrix directly using the standard summation matrix S instead of the "constraint matrix" A proposed in the paper.

This form of projection matrix allows one to compare it with the other projection matrices used in the literature; essentially ours encapsulates all the linear reconciliation methods via the term D. Moreover, it also makes it clear that the reconciled forecasts are unbiased if the base forecasts are unbiased like the standard methods.

Added a test that asserts that the two ways of computing the projection matrix are the same.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

Please tag this pr with at least one of these labels to make our release process faster: BREAKING, new feature, bug fix, other change, dev setup