Stark performance drop-off in Python

[triceo]

Discussed in #1579

In the example given in the above discussion, there is a stark difference in performance between our Python and Java solvers.

Whereas the Java solver gets to 500k/sec, the Python solver gets under a 100/sec. This is a 5000x difference.

[ge0ffrey]

I've seen cases happen that Python is pretty fast, but when one extra constraint is added, it becomes very slow.
IIRC @Christopher-Chianelli said it's related to the cut-off on how deep a method can be before the python-to-jvm bytecode transpiler falls back to interpretation?

Not sure if it's the same issue as #1579 though

[Christopher-Chianelli]

I suspect #1579 might be using Pydantic; I recall something happen in Pydantic for 3.11 or 3.12 that causes attribute access to massively slow down. Pydantic does have some C parts in; unsure if that related.