Use consistent hashing for rng seeding

Turns out the built-in `hash()` function in python doesn't behave
consistently across re-runs, and while it does return the same value
during the same program run, once the interpreter is closed and a new
instance is started, hash behaves differently.

This inconsistency is unacceptable, as the whole point of the seed is to
be able to achieve the same results consistently.

src/ui.py

@@ -1,4 +1,5 @@
 from typing import TypedDict
+import hashlib
 
 import matplotlib.pyplot as plt
 import numpy as np
@@ -127,7 +128,8 @@ class OptimizationUI(QWidget):
 
         seed = self.seed_input.text()
         if seed != "":
-            bit_gen = np.random.PCG64(abs(hash(seed)))
+            seed_int = int.from_bytes(hashlib.sha256(seed.encode("utf-8")).digest())
+            bit_gen = np.random.PCG64(seed_int)
             rng = np.random.Generator(bit_gen)
         else:
             rng = None