[clonker]

• adds ShadowLiveness: per block the phis whose shadows are live on entry and on exit
• shadows are not SSA but SSU, so their liveness is computed by path exploration
• LivenessAnalysis holds a ShadowLiveness exposes accessors to it, the stack layout generator starts consuming them in a follow-up
• the input of an upsilon now counts as used at the exit of its block only if its write is live, i.e., it is the block's last upsilon for its phi and the shadow is live on exit (before, the input of every upsilon counted)
• two assumptions:
  • reading the input at the exit requires that no upsilon sits ahead of its own phi in the phi's block, which is asserted
  • the walk asserts that every path from the entry to a phi passes an upsilon for it
• the control flow graph tests print live shadows as ^phiN[1] under LiveIn and LiveOut, and under Used if live on entry but not on exit
• the dot export prints upsilons at their position as ^phiN := value and a phi as phiN := φ(^phiN), instead of listing the upsilons as operands of the phi