Refactor buffer capacity tests to reduce duplication

[gilcu3]

Background

PR #2268 introduced buffer capacity tests across all threshold-signature protocols. Each protocol module has its own test that follows the same pattern:

1. Set up participants with Comms::with_buffer_capacity(usize::MAX)
2. Run the protocol via run_protocol()
3. Assert comms.buffer_len() matches the expected constant (with coordinator vs non-coordinator handling)

This pattern is duplicated across ~10 files:

• confidential_key_derivation/protocol.rs
• dkg.rs
• ecdsa/ot_based_ecdsa/{presign,sign}.rs
• ecdsa/ot_based_ecdsa/triples/generation.rs
• ecdsa/robust_ecdsa/{presign,sign}.rs
• frost/eddsa/sign.rs
• frost/mod.rs
• frost/redjubjub/sign.rs

Only the protocol-specific setup (building the future) differs between tests; the harness (loop over participants, run protocol, assert buffer lengths) is copy-pasted.

Acceptance Criteria

• Extract a shared test helper (e.g. in test_utils) that takes a protocol factory closure and expected buffer entries, and handles the setup/run/assert loop.
• Each protocol test calls the helper with its specific setup, eliminating the duplicated harness code.