Fix unrealistic stability for positive-feedback opamp buffer

[esaruoho]

Summary

• Adds a tiny initial offset (1e-6 V) to the opamp lastvd state in a new reset() override, breaking the metastable equilibrium that causes positive-feedback configurations to unrealistically track the input signal
• A reversed-wired opamp buffer (non-inverting input connected to output, inverting input driven by signal) should immediately rail to a supply voltage, but the Newton-Raphson solver was finding the exact metastable solution due to perfectly symmetric initial conditions
• The 1e-6 V perturbation mimics the real-world input offset voltage present in all physical opamps and is orders of magnitude too small to affect normal negative-feedback circuits

Fixes sharpie7#1009

Test plan

• Load a positive-feedback opamp buffer circuit (output wired to non-inverting input, signal on inverting input) and verify the output rails to a supply voltage instead of tracking the input
• Load a normal negative-feedback opamp buffer and verify it still tracks the input correctly
• Load existing opamp example circuits (inverting amp, non-inverting amp, etc.) and verify they behave identically
• Run the Schmitt trigger example (amp-schmitt.txt) and verify correct hysteresis behavior

Generated with Claude Code

[pfalstad]

The op amp has a number of issues but I'm very nervous about changing it because I've tried to fix them in the past and whenever I make a change, there's always one example circuit that breaks. I will take a look at this but it might be lower priority.

[esaruoho]

Understood -- no rush on this one. The change only adds a tiny random perturbation (1e-8V) to break the exact metastable equilibrium at startup. It should not affect any circuit that is not already sitting on the knife-edge. Happy to test against all the opamp example circuits if that would help you feel more confident about it.