[Math Engine] Refactor data stores & numerical linear algebra to use tensors (NumPy => PyTorch)

In order to use tensors to perform multi-variable vector calculus.

The goal is to provide calculus operators in various coordinate systems and to be able to process symbolic calculus operators into numeric computations that can be plotted.

The issue names no files, tests, or entry points. Start by locating the math engine's data stores and numerical linear algebra code, then assess how the NumPy-based structures would support tensor calculus. Done means the stores and linear algebra use tensors and can support coordinate-system calculus, symbolic-to-numeric processing, and plotting.

Written by the indexing model from the issue text.