[ViniciusCestarii]

Add 3 new generate proof test vectors for testing boundaries:

Success: a = 1, a = N-1
Failure: a = N+1 (catches implementation that does mod n on secret instead of first rejecting)

[jonatack]

cc BIP authors @andrewtoth, @RubenSomsen, and @theStack

[fametrano]

ACK 60b4c75, tested: the new vectors agree with an independent implementation.

The PR adds success cases for a=1 and a=N-1 and a failure case for a=N+1 to the proof-generation vectors. With the existing a=0 and a=N failures, both ends of the valid range are now pinned from both sides.

I ran all 14 rows of the CSV through btclib-ecc's BIP374 code, with and without libsecp256k1. Every success proof matched byte for byte and verified. The four failure rows (a=0, a=N, a=N+1, B at infinity) were refused. Re-running gen_test_vectors.py reproduced both CSV files exactly. The verification file is unchanged and already covers e=N and s=N, so I see no boundary missing from this change. Script and logs on request.

[theStack]

LGTM, can you add a small changelog entry with bumped PATCH version?

[ViniciusCestarii]

Forced-push afaa856 addressing #2335 (comment)