This reverts commit 161cc034.

- copy Bugner's C version of the solver
- compare my results with Bugner's -- good agreement
- compare against literature results (Papadopoulos 1998,1999) --
  discrepancy in sign of vr, ur, otherwise good agreement

@with_signature: only need one LineNumberNode per @accepts, @returns unpacking block (!100)

* @with_signature: only need one LineNumberNode per @accepts, @returns unpacking block

* @with_signature: Insert 'LineNumberNodes' when unpacking vectors