feat: optimise assembler pipeline

[DavePearce]

This pipeline is starting to take a non-trivial amount of time for larger modules (e.g. oob). The two main issues are:

• asm.LowerMixedMacroProgram()

• asm.Compile(nasmProgram) --- here, translateSkip() in particular is the main culprit. This could be due to branch table optimisation and/or the preprocessing step which searches forward.

UPDATE: can confirm the second issue is related to branch table optimisation. Specifically, for the very ugly reference table hacks in oob (i.e. g1_discount() and g2_discount() and also the main oob function). Use of a dynamic DCNF representation might be the only hope here.