isMaxBBWeight is ineffective; it is checking for an exact value.

Simplest fix is to make it an inequality, but we might want to think more broadly about how the jit should handle extremely large profile weights. Note we're unlikely to see these with real profile data, but we can see it with random/synthetic data.

For example, in #50743, a very deep loop nest with the default by-8 scaling per loop level eventually saturated a float.