Should munge reference designator names to be alphanumeric and unique across case insensitivity.

We should support the one line attached model ("model=" attribute) that spice-sdb has for backward compatibility.

gEDA netlisting is, in general, case sensitive while SPICE is not. This can lead to shorts in a SPICE netlist when netnames differ only in case. Or, it can lead to a SPICE simulation that's OK, while the printed circuit netlist has an open circuit due to a case typo.

Hi John! It would be very helpful to have a tutorial on slotted components
in the gnet-spice-noqsi documentation: how to work with existing symbols,
how to properly construct your own, how the backend interprets the
pinnumber and pinseq attributes, etc.

I can get it working with some symbols, but it's quite confusing, nevertheless,
see the discussion in lepton-eda issue #477.

In the following test schematics (attached), the backend works with the stock
dual-opamp-1.sym, but does not work with my opa2x_dmn.sym and Kai-Martin's opamp_dual.sym, opamp_dual_pwr.sym.

$ lepton-netlist -g space-noqsi -o - 1-test-stock.sch

* lepton-netlist -g space-noqsi -o - 1-test-stock.sch
* SPICE file generated by spice-noqsi version 20181225
* Send requests or bug reports to [email protected]
X1U1 p1 m1 vcc vee o1
X2U1 p2 m2 vcc vee o2

$ lepton-netlist -g space-noqsi -o - 2-test-dmn.sch

* lepton-netlist -g space-noqsi -o - 2-test-dmn.sch
* SPICE file generated by spice-noqsi version 20181225
* Send requests or bug reports to [email protected]
Error: U1 uses slotting. You must list its connections by pinnumber,
  not pinseq.
U1  ne5532 ne5532_ti
1 errors.

$ lepton-netlist -g space-noqsi -o - 3-test-km.sch

* lepton-netlist -g space-noqsi -o - 3-test-km.sch
* SPICE file generated by spice-noqsi version 20181225
* Send requests or bug reports to [email protected]
Error: U1 uses slotting. You must list its connections by pinnumber,
  not pinseq.
U1  TL082
1 errors.

spice-noqsi-test-slotted.tar.gz

Should probably complain when the user has IO pins at top level.

Ngspice globally aliases GND to be node 0, the datum node. This causes problems with subcircuits:

1. It is perfectly reasonable to connect a subcircuit's GND to something other than node 0.
2. A subcircuit without a GND pin will have its internal ground unconnected in a printed circuit netlist, but connected to node 0 in ngspice. Thus, it may simulate fine but not work when constructed.

In my opinion, this is an ngspice misfeature, but it appears we'll have to live with it. Therefore, we should alias GND to something ngspice won't recognize. Only in subcircuits? Configurable?