fix: correct output multipliers for all previous Azure OpenAI models + docs 7d→30d

[msftnadavbh]

Summary

Fixes the output_multiplier for all "Previous Azure OpenAI models" in the PTU sizing calculator, and updates the CLI data-collection script time window.

Changes

PTU Multiplier Fixes (ptu_models.py)

All four "Previous Azure OpenAI models" had output_multiplier = 4 — should be 3.

Model	Before	After	Verification
gpt-4o	4	3	Foundry calculator: RPM=444, in=917, out=250 → 300 PTUs (only works with 3×)
gpt-4o-mini	4	3	Official MS docs example table: RPM=1000, prompt=5000, completion=50 → 140 PTUs (only works with 3×; 4× gives 145)
o3-mini	4	3	Same "previous model" category; docs: "older models use a different ratio"
o1	4	3	Same "previous model" category; docs: "older models use a different ratio"

Sources:

• Foundry PTU calculator (gpt-4o verified against real production workload)
• MS docs latency page (gpt-4o-mini verified against official example table)
• MS docs onboarding page ("older models use a different ratio")

Docs Update (USAGE_EXAMPLES.md)

Changed the CLI data-collection script time window from 7 days → 30 days for more representative workload sizing.

Testing

All 150 tests pass (56 PTU-specific tests).