Improvements to work depth analysis

[hodelcl]

This PR adds various improvements to the work depth analysis for SDFGs. These improvements include:

• Performance improvements: Analysis now runs in minutes compared to hours (or days?) on big SDFGs. Biggest performance gain comes from optimising the Max expressions which occur from branches. Instead of Max(x + a, x + b), we now formulate it as x + Max(a, b). If x also contains a max expression, the performance gains are even bigger, since this generally allows for more Max expressions to be evaluated symbolically.
• Assumptions: It is now possible for the user to pass assumptions for the size of the symbols to the analysis. For example one can assume that N>5 and hence the analysis evaluates Max(N, 5) to N. Sympy does not support such assumptions out of the box, hence this PR includes assumptions.py, which handles all kinds of equal / greater / lesser assumptions. The user can enter the assumptions directly into the VS code extension as implemented by this other PR.
• wcr edges: Priorly, wcr edges were ignored by the analysis, which meant that reductions had work and depth of 0. Now, wcr edges are included in the analysis and a wcr edge with N writes to the same memory location adds N to the depth (and work respectively).
• Some more minor improvements.

[phschaad]

Looks very good overall, small comments and questions only.

[hodelcl]

Accidentally pushed different work to this branch, sorry! Reverted it again.

[hodelcl]

I implemented your proposed changes.

[phschaad]

LGTM, thanks for addressing the comments!