fix(prover): make WriterstoEqual actually compare both encodings

[coeniebeyers]

Fixes #3731.

WriterstoEqual serialized both operands into one reused bytes.Buffer; the buffer's backing array survives Reset(), so the comparison matched a slice against itself and any two same-length encodings compared equal. This made LoadSetup's verifying-key↔SRS check always pass, silently.

Each operand now gets its own buffer. Includes table-driven tests plus TestWriterstoEqual_Regression_SameLengthDifferentContent, which fails on the previous implementation, and a doc-comment correction ("reports whether" implied a boolean return).

Fix authored by @mkrielza.

Checklist

• I wrote new tests for my new core changes.
• I have successfully ran tests, style checker and build against my new changes locally.
• If this change is deployed to any environment (including Devnet), E2E test coverage exists or is included in this PR.
• I have informed the team of any breaking changes if there are any.

No breaking API changes.