Guard QBI phaseout division when po_length is zero (refs #8216)

[vahid-ahmadi]

Refs #8216 (84 divide warnings reported from policyengine.py).

Summary

The QBI phaseout in qbid_amount used unguarded division:

```python
reduction_rate = min_(1, (max_(0, taxinc_less_qbid - po_start)) / po_length)
```

When the configured phaseout length is zero for a filing-status / year combination this produces RuntimeWarning: invalid value encountered in divide and NaN values that propagate through the deduction.

Replaced with np.divide(excess, po_length, where=po_length > 0, out=ones):

• po_length > 0 — behaviour unchanged.
• po_length == 0 — rows above the threshold are treated as fully phased out (reduction_rate = 1), rows at/below as no reduction.

Scope

This fixes the QBI source of the 84 warnings reported in #8216. The labor-supply variables I inspected (*_behavioral_response.py, behavioral_response_measurements.py) already use guarded np.divide / np.clip / explicit masks, so they are likely not the source of the remaining warnings.

If warnings persist after this lands, running the test suite with -W error::RuntimeWarning will surface any remaining unguarded divisions.

Marked draft so reviewers can confirm:

1. The behavior I chose for po_length == 0 (fully phased out above threshold, no phaseout at/below) matches what the IRS regulation intends — alternative interpretation could be "no phaseout regardless" (reduction_rate = 0).
2. There are no parameter-tree configurations where po_length == 0 is currently exercised by tests (in which case existing tests would lock in the wrong behavior).

Test plan

• Existing test suite passes.
• pytest -W error::RuntimeWarning no longer fails on QBI variables.
• Spot-check qbid_amount for a filing-status / year where po_length > 0 to confirm no behavioral change in the common case.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (cc5bd64) to head (fc6f081).
⚠️ Report is 105 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff              @@
##             main     #8217       +/-   ##
============================================
+ Coverage   71.72%   100.00%   +28.27%     
============================================
  Files        4658         1     -4657     
  Lines       67710        50    -67660     
  Branches      341         0      -341     
============================================
- Hits        48563        50    -48513     
+ Misses      19140         0    -19140     
+ Partials        7         0        -7     

Flag	Coverage Δ
unittests	100.00% <100.00%> (+28.27%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 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.