Healthcare Readmission's Gender gap table hides that bias direction reverses between the biased and mitigated model

Where: Healthcare Readmission/README.md:31 and main README.md:80/:709 (Results at a Glance / Audit 06 table), both showing "Gender | 0.02% → 0.04% | +100% ↑".

The gap: both tables show unsigned magnitudes, which hides that the direction of the gap actually reverses between the biased and mitigated model, not just its size:

$ python3 "Healthcare Readmission/unfair.py" | grep -A1 "by Gender"
── High-Risk Flag Rate by Gender ────────────────────────
  Fairness Gap (Gender): -0.02%

$ python3 "Healthcare Readmission/fair.py" | grep -A1 "by Gender"
── High-Risk Flag Rate by Gender ────────────────────────
  New Fairness Gap (Gender): 0.04%

In the biased model, females are flagged high-risk slightly more than males (-0.02%). In the mitigated model, males are flagged slightly more than females (+0.04%) - an actual sign flip, not just a magnitude change from 0.02 to 0.04. Both tables' unsigned "0.02% → 0.04%" presentation reads as a straightforward (if small) increase, when the reality is the disadvantaged side switched.

This repo's own table already has a precedent for showing direction when it matters - Benefits Denial's Age row is shown signed ("-2.72% → -2.79%") specifically for this reason.

Fix direction: show Gender signed ("-0.02% → +0.04%") in both READMEs for consistency with the Age row's precedent, or add an explicit note about the direction reversal. Worth noting in the same breath that both gaps are near-zero and not the headline finding of this audit (race/age are).

Update Healthcare Readmission/README.md:31 and main README.md:80/:709, using the Benefits Denial Age row as the formatting precedent. Run the unfair.py and fair.py commands shown in the issue to verify the signed gender gaps, then confirm both tables show the direction reversal or explicitly explain it.