Let compile_train take extra updates

[jessegrabowski]

compile_train couldn't write state that no gradient produces — a target-network sync, a Polyak average, replay priorities. The obvious workaround, compile_kwargs={"updates": ...}, hit a TypeError naming an internal function.

Now there's an extra_updates argument, and an updates key in compile_kwargs is taken as one. Either spelling raises if it writes a variable the step already writes — a target sync quietly overwriting an optimizer buffer would leave the rule configured but doing nothing.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.42%. Comparing base (cf70a80) to head (7b3e308).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #76      +/-   ##
==========================================
+ Coverage   95.39%   95.42%   +0.02%     
==========================================
  Files          45       45              
  Lines        1782     1792      +10     
==========================================
+ Hits         1700     1710      +10     
  Misses         82       82              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.