Bugfix: finetune.py from collections import OrderedDict

[byarbrough]

Following the instructions in the README.md gives

Maximum ROA verified for 100 consecutive iterations. Training complete.

Training complete. Final verified target ROA: 0.9900
Time Used for Fintuning: 372.71757984161377
Traceback (most recent call last):
  File "/app/src/finetune.py", line 462, in <module>
    filtered_state = OrderedDict()
                     ^^^^^^^^^^^
NameError: name 'OrderedDict' is not defined

This is confirmed with pyright:

/src/finetune.py:462:26 - error: "OrderedDict" is not defined (reportUndefinedVariable)

The fix is a simple from collections import OrderedDict as in the other files.