Add physical interpretation of the concept of a Dimension

Copilots comment in PR #112 in [this comment](https://github.com/atmoos/Quantities/pull/112#discussion_r3825159767).

On the dimensionless dictionary contained in a class in Arithmetics.cs

> Dimensionless cancellation is keyed by the exact measure wrapper, so compatible angle units do not cancel when their representations differ. For example, dividing an angle measured in `Metric` by an angular velocity measured in `Si/s` leaves a `°/rad` factor in the resulting `Time` measure instead of producing seconds (the polynomial already carries the conversion factor). Normalise compatible named dimensionless measures before applying them rather than accumulating each wrapper `Type` independently.

We should fix that 😉

Start in Arithmetics.cs and read the dimensionless dictionary logic described in the PR #112 comment. Reproduce the Metric divided by Si/s case, then verify that compatible named dimensionless measures cancel and the resulting measure is seconds rather than °/rad.

Written by the indexing model from the issue text.