docs: correct the n-bound rationale — no USDC was ever at risk

[VickyXAI]

0.7.2 shipped a claim that isn't true. Correcting the record.

What was wrong

The changelog said an out-of-range n "took payment, then 400'd at the provider and lost the prepaid USDC."

BlockRun's gateways settle on success — settlePaymentWithRetry runs after the upstream call — so a request that dies at the provider settles nothing and costs the caller nothing. I inherited the claim from a reviewer and from the t2i schema's own comment, and repeated it without checking the settle ordering.

That's the same mistake the SDK's "API error after payment" wording keeps producing: it reads as funds gone on failures that took nothing, and it has now misled two reviewers and me (fixed in BlockRunAI/blockrun-llm#25).

What stays

The n bound. It saves a pointless signed round-trip, a facilitator verify, and an upstream call, and answers 400 locally instead of surfacing the provider's error late. It just was never about losing money — and a changelog that cries wolf about funds is worse than one that says nothing.

Scope

Docs and comments only; no behavior change. 277 tests pass.

PyPI renders README.md rather than CHANGELOG.md, so the published 0.7.2 description never carried the claim — no republish needed. The corrected text ships with the next release.

Related

• fix(errors): stop claiming payment was taken when nothing settled blockrun-llm#25 — SDK reports whether settlement actually happened, instead of "after payment" on every failure
• BlockRunAI/blockrun#249 / BlockRunAI/blockrun-sol#50 — gateways validate image bytes before the 402, so garbage input 400s instead of 500ing after a verify