Keystore signer impl

[connorwstein]

Note I thought about passing a hash function parameter (e.g. nilable crypto.Hash, with the idea that the more we can encapsulate cryptographic stuff the better) but ended up thinking its not worth it:

• Different signature schemes have different hashing approaches, for example ECDSA libraries generally just accept a 32 byte digest only and expect the caller to hash whereas Ed25519 by default allows arbitrary message lengths (internally uses SHA-512). Existing code generally uses the schemes in that manner, so leaving it this way is less disruptive to existing code and less likely that clients accidentally hash something twice or error for not hashing on ECDSA.
• Took a peek at BTC for example which also uses ECDSA on S256 but unlike EVM chains it uses a double SHA-256 hash of a serialized BTC transaction. Supporting a double hash would complicate the API further, so again seems better to leave that client side.

[github-actions]

✅ API Diff Results - No breaking changes

---

📄 View full apidiff report | 📚 Learn about apidiff

[connorwstein]

General pattern I'm thinking about here:

• KeyName is bound to private key material specifically and only used in APIs which need that access
• For APIs that don't use private key material directly (but still belong in the API as they do cryptographic operations), supply all the inputs up front so its stateless. More flexible that way

[connorwstein]

critical issue caught in testing the signing, casting instead of properly converting.