Fix PyPy build failures: handle float overflow without OverflowError

PyPy follows IEEE 754 strictly and produces inf instead of raising
OverflowError on float overflow. Add an isfinite() guard after the
Izhikevich neuron voltage update so the overflow reset works on both
CPython and PyPy.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: CodeReclaimers

neat/iznn/__init__.py

@@ -9,6 +9,8 @@
 http://www.izhikevich.org/publications/spikes.pdf
 """
 
+from math import isfinite
+
 from neat.attributes import FloatAttribute
 from neat.genes import BaseGene, DefaultConnectionGene
 from neat.genome import DefaultGenomeConfig, DefaultGenome
@@ -109,6 +111,12 @@ def advance(self, dt_msec):
             self.v = self.c
             self.u = self.b * self.v
 
+        # PyPy (and other runtimes) may produce inf/nan instead of raising
+        # OverflowError.  Apply the same reset in that case.
+        if not isfinite(self.v) or not isfinite(self.u):
+            self.v = self.c
+            self.u = self.b * self.v
+
         self.fired = 0.0
         if self.v > 30.0:
             # Output spike and reset.