merge boundary constraints into AlgebraicGraph and support variables

[tohrnii]

Variable support to transition constraint was added in #100. We should also add support for variables to boundary constraints. It might be useful to also add boundary constraints to the constraints graph in the same PR we add the variables.

The original issue with syntax is described in #61.

[grjte]

It now looks like we should do this task before completing #56, since having the boundary constraints in the constraints graph will simplify GCE generation a lot

In addition, to avoid rebuilding a lot of the symbol table information during the GCE codegen, we should consider some other adjustments to the IR:

• accumulate the inline constants while building the graph. We can use this info in the future for optimizations, and we can use it directly in the GCE codegen instead of accumulating them there
• provide access to the identifiers. A subset of this information is required during the GCE codegen. To avoid rebuilding it, we can either provide read access to all of the identifiers or we could provide just the subset we need, but since that's more work I'm inclined to just make the identifiers readable, since it's the simplest approach. We can think about readjustments later.

[grjte]

Closed by #116 #117 #126 and #160